Tech Talk # 8

Tech Talk # 8

Show Video

e e e e e e e e real for hello everyone am I audible yes St yes so very good evening everyone this is Dean Singh from esot 69 Calum Alliance University so uh yeah today's today's and exciting evening we are going to start with our Tech talk so uh I'm thrilled to welcome you all to this exciting session where we will explore how technology is transforming the world of Finance in the ways we never imagined Our Guest is like our guest is Mr a it's my privilege to host and introduce Our Guest uh he's a season Tech expert with over eight years of experience in software engineering AB is currently a senior engineer at marketfield where he's helping investors unlock their uh extra income and through automated trading Solutions and before joining Market feed I I hope you all must have heard about Reserve pay right it's one of the like biggest Payment Solutions where you we use like Google pen phone pay we it redirects us to reserve pay that's how we pay and he worked there as senior engineer okay and he contributed there as you know creating a scalable payment system and before joining Reserve pay he was also he was also working with Zolo and CH Street where he let projects ranging from gamification systems to cloud storage solutions and API governance so let's get ready to learn and ask questions I hope this everyone learns from this session so please give a warm round of Welcome to Mr hello I hope I'm audible yeah you are audible welcome thank you thank you so much D for the introduction and yeah so U I hope to share certain learnings that I had from my previous experiences and show you an introduction into the fch world uh looking forward to uh your questions and yeah let's get started okay I'll be sharing my screen just give me one moment uh till the time abai is sharing his screen let me tell about the how the things will go today abai will explain about the know presentation and the topics and after that we'll have a Q&A session I hope you guys focus on the learning part then you can ask your doubts to a yes A we can see your screen all right let's started so uh fintech is a pretty broad domain and technology in fintech is even broader so uh it is very hard to con condense all the knowledge in 45 minutes so I'd like to uh slightly go a bit faster in this presentation I hope all of you keep up with me I try to keep a bit more Focus because there's a lot to unpack okay let's get started so uh as introduced by Dean so I worked as a staff engineer at Razer pay for uh 3 years and now I'm currently uh leading the trade automation team in Market field so topics for today uh we are looking at the current landscape of ftech in India uh how engineering is done at razor p and how engineering is being done at Market field and the future of ftech what are the things that we can look look look ahead for and the Q&A uh sorry was there a raised hand uh anything uh is there an issue no no I think we should go ahead with the session all right okay so uh let's get into the first section which is ftech landscape in India so I'd like to uh start the session with a poll so uh I hope this particular slido is visible for you you can scan the QR code or you go to slider.com and enter that particular code to be able to participate in the poll um take one minute to go through the question and you can select your uh select your option which that you think is correct so the question is which country processes highest amount of payments digitally uh they want let me know if there is an issue with the pool uh yeah okay yeah I can see the polls coming in lot of people voting for India and there are few people who voted for United States okay cool a lot of polls coming in for India almost 90% of people think it's India is the highest and now it's cross 91% all right so I think we've reached a certain stage where I can see okay yeah now I see that people are selecting China as well around 7% votes for China okay so there are people who think it's India obviously okay cool I think now uh we can stop the poll I guess uh anyway it's very clear that majority of you feel that India processes highest amount of payments digitally I'm stopping the pool now so the answer actually is China uh while India is a growing economy with a lot of people participating in digital payments we are nowhere close to China China processes close to 9.3 trillion doar of payments digitally India somewhere close to 850 billion and we are expected to touch a trillion so we are nowhere close uh compared to China but we are almost there USA and UK are not at par with the Asian countries when it comes to digital payments primarily because of the lower amount of population as well as their volumes are lower but their uh amounts are higher all right so uh looking at the current uh ftech Market in India so uh if you have noticed uh since 2020 there has been a huge investment scoring into uh ftech domain especially from the US into the country uh most of the Investments are focused on the SAS products and new banking products but recently we have seen a greater push towards insurance and a greater push towards lending as well as payments have more or less been the same so we have seen almost 2x growth since 2022 and uh in 2025 is expected to cross $1 trillion Mark which I was saying before um if you look at uh certain domains in fch so you have payments which is the most obvious and what people think ftech is but there are a lot of work that is being done in lending in insurance and investment products new Banking and as well as crypto crypto is huge uh before uh there are some regulatory action that is taken against them but now we are seeing again a slow boom uh since uh the rise in the Bitcoin value so one of the things that is making uh fintech so popular and growing at the rate that which we are seeing is primarily due to India stack so when it comes to finance one of the things that's most important is identity and security being able to understand who is the person that you making a payment to or who's the person that you're giving a loan to is very important and especially looking at their credit history so AAR and digilocker being one of the core elements of any fintech enables them to acquire customers faster and onver them into their products much easier so um if you have used any of if you have opened a bank account or used any fch product the first thing you do is do your kyc using other and you would submit your proofs using dig loer and most of the companies and products that are built are on top of the existing stack like UPI and rupe and we are also seeing a rise in open networks so people who don't know open networks these are fairly new Concepts in the market so OK for example is open Credit enabled Network so if you are a person who wants to go take a loan uh for buying a car for example you go to a bank and take a loan but if you are a person who is a shopkeeper who wants to take a loan to fix an AC for 2,000 Rupees now where would you get that so those sort of micro loans uh are very hard for Bank to disperse because the cost of issuing a loan is very high because you have to do uh the kyc you have to do the credit checks and all that Oken enables anyone to participate in an open network where people can come in and lend micro loans and the verific and there can be a lot of digital Network players who can participate in enabling things like verification and credit checks and all that so onc is another example uh if you have not heard of onc it's basically open network for digital Commerce again enabling a democratized digital platform for Commerce uh is the main focus for onc things like namaya 3 in Bangalore for example is a sort of an app which is built on top of ondc uh having these fundamental layers which enables Innovation to happen is the key to why there is a rise in the fex space in India so uh now let's get into the the core of it which is engineering aspects so I'd like to talk about uh certain engineering things uh how engineering is actually done at Razer p uh to start with let's get into another poll I hope you're able to see this poll as well you can again scan the QR code or you can go to slido and enter the code to particip B in the poll yeah so here uh the question is how many orders would have been processed by Razer pay in the time it took for you to finish reading this question okay polls are coming in uh there are people who think it is 10,000 orders uh per second or 10,000 orders within the time you took to read the question okay there are people who think it's one lakh okay so uh to repeat the question it is basically how many orders would have been processed by R in the time it took for you to finish reading this question so if you took around 2 seconds or 3 seconds to read the question how long how many orders would have been processed by Raz in that time okay there are people who I think there are thousand orders would have been processed cool I think we have reached a certain threshold okay uh let me stop the poll and let's see what is the right answer so the right answer is uh 10,000 ERS so it's approximately 10,000 1 lakh is too much I don't think any country processes one L orders within few seconds but 10,000 is a pretty high number yeah so now let's get into uh a use case right let's say you are a person who's building a product uh you're building an application so how would you get started usually the typical scenario is you have something like a you have a database you have a full stack application which includes it could be a mon stack or it could be a mean stack you would have your front end back end and everything amalg into a single app and this application can largely handle around th000 requests per second at Max and you'll be able to decently have a launch but if you have a capacity where now you saw the previous L rer pay processes around $10,000 per second so imagine you getting thousands of requests from different different people you can start to see certain pressure on your app like your app will not be able to scale and your database will not be able to scale because it's a single app and a single datab now imagine you have people who are attacking your application there are hackers who want to hack into your network and steal your data uh things like credit card information or bank account details or worst case scenario steal people's money so you you your application is now even under more pressure now think there are 20 or 30 or even 100 developers who are continuously adding new features into your application now it is impossible for the application to hold because there are continuous changes being integrated there are continuous requests being uh processed and there are continuous attacks being handled now one single application would not be able to handle so much and imagine now you have a lot of analysts who are analyzing the data that is being stored in database your database also go goes for a test so the setup that you would ideally think how an application or a system would work would not scale uh when you are dealing with this large volumes so when we look at razor pay scale right uh per second at at Peak Razer pay processes somewhere close to 3,000 transactions these are especially true during the times of IPL where we see highest amount of orders being placed most of the orders come from dream 11 s or uh certain which are the products that are mostly consumed by people during the time of IPL U and during that time what we notice is we ensure that we have 99.99% of up time that means most of your requests will be successful and in a second we process close to 40,000 uh requests so this request not necessarily be orders it could be uh read requests account statements logins any kind of request into the application to handle all of this uh as I mentioned earlier a single application would not be able to hold and a single database would not be able to hold you need to have a distributed stack where each and every service which is responsible for a particular fun function is carved out so we have close to 200 plus services that are present in razor and somewhere close to 1500 deployments being done as you seen in the earlier slide there are hundreds of developers who are continuously pushing changes and there are thousands of deployments that are done in a month so if you look at the department structure of Razer pay the way it is structured is divided into three layers there is a product layer which is on top then there is a platform layer and there is infrastructure layer so the product layer is what most most people refer to when they think about Reser period payments payouts Neo banking payroll things like that but actually that is just one third of Razor pay the the remaining two3 is actually the platform layers and the infrastructure layers if you look at the platform layers here we uh the platform layers contain services and systems which are commonly used uh by all other product features things like absorbability which includes monitoring the systems and Ure inreasing the systems common notification Services managing the AP Gateway which includes uh threat production and then accounting pricing and reconation systems which are largely common for all the products and then one more level down then you have these horizontal teams like devops data engineering and security which ensures that the platform and the product are completely secure so if you think about the structure product's job is to develop the product platform and infrastructure team's job is to ensure that the product development is done faster in a robust Manner and a safe manner so this is what differentiates between a fintech and a regular company in fintech there is a lot of investment going into the platform and infrastructure side along with the product whereas in a regular Product Company the most focus is on product and platform infrastructure takes a little bit of a backseat and the area that I've highlighted the accounting pricing and reconation that is a team that I worked and led there so uh as I mentioned in the previous diagram if you look at the architecture uh you have different features on the left side uh different dashboards and then you have different databases on top which include my SQL redis elastic search post Centra and then on the bottom you see lot of different microservices present so each and every microservice utilizes a wide variety of databases and interact with wide VAR of systems so it's a completely distributed system and uh one of the things I mentioned uh in the previous diagram was a single database would not be an IDE choice for serving the entire company in this case in razor pay as youve seen in the previous slide we have so many different databases each database has a particular purpose if you look at myql and postas the purpose is to store the entities and uh to store the transactions so here critical things like accounts balances and the financial transactions are stored these are largely relational databases and uh the most important thing here is having aset principles present redis is used for storing things that where you need quick access things like configurations distributed locks or you know quick quick access data for caching purposes and elastic search is used for enabling search so to give you a sort of an example if you are a customer of Razor pay if you a merchant who wants to log into the application when you are receiving a payment your payment is largely stored in myql and your accounts and balance are stored in myol or post your bank statement is actually powered through elastic search where you would be able to search through the bank statements using a particular keywords your entire configuration and your setup would probably be stored in redus so this is how different purposes Drive different database choices here and for each product when we choose what database is needed we would think about what is the purpose and we take a call but if you look at this uh all these data B are mostly geared towards product none of them actually serve the analytical purposes when you say analytical purposes the purposes where you are doing big data analysis you're trying to derive insights out of the data or you're trying to generate reports on last three months or six months or things like that normal databases will not be able to support uh these analytical features at this scale so which is where uh the differentiation of oldp versus oap comes in oldp databases are transactional databases and oap databases are analytical databases so on the left side what you seeing here are MySQL and postris which are transactional databases and on the right side the databases that you see which are TB elastic search and S3 these are analytical data stores or analytical Data Systems to transform the data from the left to the right which is from a transactional data system to an analytical data system requires some we of processing and this is what we call big data processing in big data processing in razor pay we use Kafka and the paches park there are lot of other tools but these are the primary tools that are responsible for this transformation so here we do ETL things around uh taking the data breaking it down normalizing it or denormalizing it joining it with other databases and then creating a final uh final sort of a views that can be consumed for reporting purposes or for analytical purposes so TB is a very new age database which is actually built on top of post but it's geared and tailor made for analysis elastic search is primarily for search and S3 is stored for recording all the reports that are present there so uh some of the technologies that are play in analysis so as I mentioned in oap olp toap replication you have Maxwell deum these are certain tools that can be used to transform the data to replicate the data from your my SQL and postgress to to Kafka and tidb as I mentioned S3 is for file stores and tro is another database engine that is used to read the data on top of the data LS and apach Spark primarily used for distributed processing so this itself is a pretty expensive and pretty complex process and there is like a dedicated team which is working on focusing on transforming this data uh from a storage friendly system into a reader friendly system when it's a storage friendly system if you're storing data in LP the purpose is to write faster you would want to as soon as a transaction is to be processed it should be recorded as soon as possible so we're talking about latencies like uh thousands of transactions in 60 milliseconds things like that but when it comes to analysis you would want to read large volumes of data excuse me yeah so for uh sorry just give me one second yeah so far we spoke about product development platforms analystics but we never spoke about security being a ftech company security is of Paramount for razor pay and it is not just one level but there are three levels of security present in razor pay so if you look at it these are the three levels the first level is VA which is short form for web application firewall so this faav generally is responsible for prevention of DS attacks or datos attacks are essentially denal of service attacks that can be targeted by malicious players on the internet to take your network down V ensures that these DS attacks are prevented not just DS attacks attacks like cross-side scripting uh cookie overloading session takeovers anything that is preventable something that is predictable through certain patterns are prevented by W and the second layer of defense is application load balancer so this load balancer essentially ensures that the system is not overloaded it distributes the load across the systems evenly and it also has certain rules if someone is trying to access your application without login then application load balancer prevents it before even that happens and application load balancer can also have certain rules that are written for example IP restriction if you don't want your application to be accessed by some other country for example then you can have this entire country blocked at the AL level then the third layer of production is Edge Edge is uh internal terminology for API Gateway Kong is the technology that rer uses for their AP G and Edge is largely responsible for your authentication and authorization so who is accessing this information that is uh the role for authentication do they have access to uh read the data that is what authorization does so Edge is largely responsible for handling the authentication and authorization of it so in order for anyone to be able to access razor pay systems they would have to cross all the three levels of defenses and only then they would be able to actually perform a payment or check out the orders or anything like that and uh this is again uh something that is unique for fintech because this three levels of protection is something that you rarely see in a regular product but uh in finance it's important to protect users data and especially money so that's why the more number of layers of security the better but it always comes at a cost because the more layers you add to security uh you increase the latency uh and the travel time for your request and also points of failure so it's also something that's a pretty balancing act that needs to be taken care of uh so this is largely the text T that is used by Razer Bay all the things that I mentioned earlier are sort of collated in this one particular place uh so mostly systems are return in PHP Python and Goan and you have your databases like my SQL and RDS and devops uses these tools for maintaining the infrastructure now uh I've spoken so far about razor pay I'd also like to talk about my uh latest gig at Market VI so Market feed is essentially uh a product which enables automated trading in user accounts so it's a pretty highrisk product if you look at the risk apped it is higher than your Equity mutual funds but the returns are also higher so I'd like to take a a moment to present another poll here yes so yeah this is a poll you can again scan your QR code or go to slide and enter the code to access the poll so the question is what is the highest amount of of loss faced by an automated trading platform okay the answers are coming in uh there are so people are voting for all options 1 million 10 million 500 million a billion dollars interesting okay so I'll stop the poll here so the right answer is $500 million and this loss happened in a span of just 45 minutes so there was a developer who wrote some bad code and deployed it into the system without knowing it and then uh it started started placing a lot of orders into the market and eventually $490 million got wiped out in the span of 45 minutes before the Traders could actually shut down the application so this happened uh in the company called night Capital it's a wealth management company in us and they had to eventually liquidate their company and sell themselves to their competitor after this issue happened so this tells you that trading is a pretty a risky product and it requires a lot more atten attention to detail and a lot more focus when it comes to managing the systems and services so what is market feed so Market feed is a platform that helps you ear extra money through automated trading so uh the thing unique about Market feed is it connects to users brokerage accounts and automates trades so there's no money that is actually being taken into in any product that you are usually uh working with you would actually have to give your money and then the people would actually make money for you but Market feed we don't touch users money we only uh automate trades in their accounts so the complete control is with the users and they have complete transparency into what's happening into their account so uh historically uh we have delivered close to 22% returns uh at the highest and on average people uh got somewhere around 16.43% returns uh in the past few months and we manage close to 226 crores of EDM so we only started 2 years ago and this is the current AUM uh that we are managing and we are a pretty small team so if you think about it the total team size is around 50 people and the engineering team size is 15 members and uh 15 members are essentially Building Systems and services to trade with hundreds of crores of capital so this is the high level architecture of any sort of algorithmic trading platform if you look at it so on the left side you have the market data which is essentially uh the ongoing ups and downs of the market the charts that you would usually see in as like Zera or in any other stock market charts so this is Market data which includes the price Ms of different uh stocks and options this Market data is funneled into a sort of a pipeline which uh in our case we use mqtt protocol for people who are familiar with mity protocol you would have noticed that it is actually used in iot uh where devices with when they have to talk to each other they use mqtt protocol we use mqt protocol to stream Market data because that is the fastest way to actually communicate information uh which is know uh which is sort of a bit by bit in nature so we have three levels of systems so one is a Quant engine which essentially runs algorithms by reading the market data and takes decisions so these decisions are called trade signals and these trade signals are Ved through a risk management system which ensures that whatever that we saw earlier right where a mistake happened and then dummy trades were placed which led to huge amount of loss such scenario doesn't happen so these risk management systems are developed to prevent any anomalies that can potentially hampers users money so once a trade signal passes through the risk management system it goes into a trade automation platform which is responsible for placing orders or ensuring that the orders are executed these orders are essentially uh the trades that are placed in the market by the broker in our case we partner with IFL to execute the broker to execute the orders and these orders are placed in the actual Stock Exchange like n or BC uh to give you context uh Market feed is able to process thousands of orders which are valued at hundreds of crores but they are executed in a blink of an eye uh the moment it took you to read this entire thing you would have probably placed thousands of orders and moved hundreds of clores of money uh into buys and sells so this is the pace at which we're talking about when it comes to algorithmic trading so creating algorithms for trading is a pretty complex process it involves machine learning and also a little bit of AI so we use uh two sort of machine learning algorithms one is prediction algorithms and the second is optimization algorithms prediction algorithms are responsible for predicting the market Ms so in our case to use random Forest regression which is basically an evolutionary algorithm so these algorithms mimic how Evolution happens so you have natural selection process where the survival of the fittest is present so that's what we use and we also use genetic algorithms for optimization so for people who are wondering how uh data science ML and AI plays a huge role in finance uh this is how so these are used for risk mitigation and recognition all these things it is used so using the machine learning techniques we develop the algorithms and we use a lot of statistical indicators to analyze the data and finally we use the distributed cloud computing to actually simulate uh the future what Market could look like to actually place the trads so the text for Market feed is fairly simple we use a lot of goang primarily because go is one of the fastest languages for computing parallely uh but also ensuring that there is contextual awareness so in parallel Computing it is fairly easy to create lot of threads and then execute jobs parallely but it is very hard to execute jobs parallely with contextual awareness where one thread knows what the other thread is doing and goang is sort of very good for that kind of a purpose so that's why go is used in fact in razip as well as in Market field I primarily worked on goang and I'll probably be working in go for at least four stable future life language which is much more suited for uh ftech domain python is primarily used for machine learning purposes for uh developing these algorithms and uh we use wide vary of databases to store user data and process information so uh this brings us to the final segment of the talk which is around future of ftech what is the future looks like so uh there is a lot more focus on sustainability and Alternate investment platforms and blockchain uh these are the three areas that I foresee as a places where more amount of disruption can take place in fintech when it comes to sustainability we are already seeing uh things like digital solar so there are companies which are able to set up a solar plant somewhere remotely and give you a fractional share and give you credits so it's like you are don't you don't have to set up a solar panel for your house but you can actually buy into a piece of a solar farm and take the fractional share of the energy credits that you get and you can use that to repay your electricity bill or for your charging station bills so these sort of solutions are originally not possible if there was no fintech but now because uh we are able to digitally move money around and digitally move credits around this is possible we also seeing a huge rise of alternative investment platforms so uh if you look at uh the historical performance of alternative investment products they have actually outperformed traditional products so this graph here is showing a difference between S&P 500 which is an index of United States uh being outperformed by the alternative investment portfolios so this alternative investment portfolios can be anything like a Reit or fractional share into art fractional share into real estates and insurance Tech products or anything like that so uh these are the books that I recommend for people who want to actually get into fintech while these are standard for all domains so if you have a good understanding of these Concepts these are pretty good anywhere but if you want to have a a thriving career as a engineer in fch I think these are a must reads clean code is very important for you to be able to write code which is scalable where you can collaborate with thousands of Engineers but without any friction and being able to produce a very qualitative and functional code designing data and Deno applications is another book that is must read for people who want to build applications which scale into know thousands of requests that you're processing within few milliseconds now uh this while you are reading clean code you would be learning how to write good code but when you're reading uh designing data intensive applications you're essentially learning about how to make that code work at Large Scale how to make it distributed how to make it consistent devops handbook is another book uh which I highly recommend not just for devops engineers but even for developers because that actually changes your thought process on software development it teaches you how to write code faster how to ship things faster in a much more safer and reliable way and finally domain driv design which is a general uh study on how to structure your data how to structure your services and systems uh well so uh if you want to know more about ftech or you would want to uh get on some mentorship or you want to connect with me uh so you can scan this QR code and uh you'll be able to reach out to my LinkedIn profile uh I'm happy to answer any questions you have and then I'm happy to chat or help you with any problems that you're facing all right so I done with the session and I'm open to taking any questions that you have I'll stop sharing my screen now and then uh you can let me know I think Dean you can take over and then let me know what are the questions that are coming thank you so much that was a lot of knowledge and am I audible right yes yes yeah that was a lot of knowledge so I will be starting with uh before moving to the question part I want you to elaborate more about your current market like your current working position that is in Market feied I want to know about the working of the market feed okay how it works like like uh I I personally invest so I use uh suppose I use grow so I place order it get placed so how your automated training program you know comes there so I know you already talked about you have uh iifl this broker video so how it works can we elaborate more before moving to the question part sure so it's a great question so for example you're using grow as an app grow is a broker as well so you are essentially using their app to place orders now uh as a person you're a student and there are people who are working on full-time jobs they don't have time to do trading and most people don't even have knowledge on how to trade as per the recent survey conducted by sebi we know that more than 95% of users actually lose money in trading especially in in options trading people lose more than 99% people lose money the reason people lose money is because trades are taken through emotion and not through fundamental analysis and research so what we essentially do is we want to automate the decision- making process on choosing which stock to pick when to buy and to sell and make it automated so that you don't have to worry about it you can go on to study your class or uh do your job and you don't have to worry the trades are placed automatically in your bage account you can open your app anytime and you can see what's being done and at the end of the cycle you'll be able to see what is the profit or loss that you would need so we primarily trade in options uh Nifty options on weekly expiries so we have a weekly cycle so at the start of the trade cycle you would be investing your money at the end of the trade cycle all the money is now recovered and whatever is the money that you end up with whether it is higher or lower will tell you whether it is profit or loss thank you so much for your explanation so we are moving to the questions part everyone if you have any doubts regarding this you can ask good evening sir prash de side I have this question for you someone asked for someone aspiring to build a career in this tech industry what do you consider the most promising fields of study like AML or data science or what what kind of well tech industry in general is a pretty broad term right and there are a lot of things at play as I showed uh in the slides as well it is not just product engineering you have platform engineering you have infrastructure there are different different domains and departments present there uh what I would suggest for anyone as a beginner who is getting into it is not to choose a particular specialization rather uh focus on the fundamentals on writing clean code good code scalable code being very good at problem solving and eventually the domain will choose you uh you will start exploring different roles when you move into across different companies and eventually you will find the fit that you like the most in my own personal career I was a full stack developer for four years devops engineer for 2 years and then a backend engineer for for three years now I a manager so I've been transitioning across different domains and experiencing different Technologies and it's important to have this sort of a diversity On Your Arsenal to be able to choose a specialization you have a long career ahead you work for up to 60 years so you have like 40 45 years of career ahead so you don't have to choose one right now thank you sir thank you okay so so uh I I already have a list of questions uh that already folks wanted to ask so uh the first one is uh as we already know that you have a strong background in areas like M Stack microservices and devops what motivated you to transition it into know fintech sector especially with the reserve pay like do you have any interest or something like this do you want to tell us no no there was no specific interest they just paid a lot of money so f a lot so I appreciate that and uh the second question is uh what are the key trends in fintech Industry that you are seeing anywh India is India is growing right so in these days a lot of fintech startups are coming so what do you see the future of fintech I think uh people who are building products on open networks would be probably the future of Tech in India as we have seen companies like rer pay pay phone pay and so many other PTM so many other companies which have built on top of UPI and rupe uh now you are we started seeing more companies like full money and Magic Pen so many other companies which are building now products on top of the open networks and new initiatives by the government so uh it's very important to keep a watch of the India stack and the regulatory Norms especially in finance bace uh and first M always gets advantage in our case uh we are seeing a new regulations coming in for algorithmic trading by sebi so people who are already into algo trading space would have a huge Advantage because in a regulated market you actually can uh capture the market a lot more faster and establish yourself as an industry leader so yeah I think I would say read a lot and follow uh follow what's happening in the in this area and then You' be able to know more yeah I think there is a hand raised yeah yeah so I have he some foreign people coming to India and they seeing that our up transaction through QR codes they find out something futuristic so why what do you think [Music] like uh okay I think your audio stopped but if I your question yeah why why do you think that some Europe some other countries wouldn't implement the same thing like UPA uh in they banging things okay so um if you look at why UPI is successful in India it is primarily because India is a cash based economy before if you had to purchase something there is more likely chance that you used cash I'm not sure how many of you folks have used cash because U you are fairly young I'm fairly old so uh when I was in college I used to pay uh in cash for my college tution fees and then even for buying tickets or anything we always used cash because online payments were pretty expensive and they were pretty unreliable so what has happened here is UPI has brought in these two things uh solve these two things one is replace cash payments easily uh for everyone which is just by having a QR code and a mobile phone and second thing is they ensured the success rate of UPI transactions are high when UPI started if we would have seen even 1% of transactions failing people would not have adopted it so much because UPI has a higher success rate uh the adoption has been higher and because we are a cash based economy even a merchant who is selling fruits or vegetables when they get money through UPI they don't think uh the money actually they don't think of it as any different from cash and they don't have to necessarily own an electronic device like a card machine or anything like that to actually receive payments so it's fairly easy to uh spread that technology especially into Tire 2 and tire three segments now when it comes to countries like Europe and us they are largely card based economy most of the transactions happens on the card networks like MX and you know visa and MasterCard so now what happens in those uh sort of segments is uh because they're not so reliant on cash uh penetration of Technologies like UPI may not happen that's why we are seeing more interest in UPI especially from countries like Bangladesh Sri Lanka Nepal Bhutan Thailand Japan South Korea which are hugely cash based economies so this technology is favorable for this sort of a setup maybe we see more action on UPI from African countries as well which are also hugely cash based economies but in Europe and UK and us I'm not pretty sure uh it makes much difference there guys if you still have questions you can continue asking yeah Adit so hello your talk was very interesting I got to know a lot of things so I have two questions from here so uh my first question would be how to think about building scalable systems when I uh say thinking about building the scalable systems whenever we develop something uh very firstly uh we do not develop it thinking in mind that we it will be delivered to let's say a million customers uh so I wanted to know that how should what should be the thinking process is is it should be that we are building for a million customers or we we are building for our own or maybe our friends and then we will scale it uh as as things happen because when we think of a scale we design the things differently we design databases we design things differently so how should that happen and my another question is uh uh there are many technologies that are appearing in Daily like not daily I should say very frequently let's say you are using redis as a database and tomorrow some another technology comes that maybe that is maybe faster uh than red that cashes things faster than R so how would you transition to that like uh what would be the thought process of your company and what would be the decision making process of your company uh how do you assess things that are coming frequently new technologies to trans transition towards them okay got so let's start with the first question right which is how do you design your systems as per the scale so uh as with any company or any startup right you start with you start with an MVP which is you know a poorly made product where you're trying to find product Market fit you already have an understanding of your skill if you're building something in fch you probably would not be building something that someone has already built ages ago because you would not be able to capture the market fairly easily so you must be building something unique something no one has thought of before a unique service you would have to find your Niche segment there you can still have scale there is no doubt about it but uh it's about identifying that size of the target segment and how many users you are targeting if you the SAS product scale is already determined right because you can understand how many people are opening your accounts how many people are subscribing to your SAS license based on that you can have an estimate on the scale that you having when you start you'll probably Target 100 accounts to open in the first month month so you can understand okay if there are 100 people using it what should be the database configuration what should be the application configuration you can plan for that when you're acquiring more users when you're growing aggressively uh then you would always have an anticipation of what would be the prodct predicted scale that could come in there are always moments where you might not be able to uh predict the scale and that's when Auto scaling comes in so if you have written your systems in such a manner that uh the autoscaling policies can actually multiply your databases and multiply your applications uh servers easily then you need not actually think too much about it uh I would always advise this for anyone whether you are building an MVP or whether you're building for 10 users or thousand users write your code in such a way that it can be autoscaled there are systems out there especially with Cloud Auto scaling is a piece of cake now as long as your system design is in line with Autos scaling you would fairly you would fairly be okay and even uh when it comes to database designs while the capacity of your database is provisioned only for say 100 users but the design of it should be done in such a way that it can be scaled for a million users if you can do that from day one you would have fairly the journey towards growth compared to uh certain place certain companies where you would have to hire in experts to refactor the entire code or redesign entire database which is a fairly expensive thing to so that's for question one question two you asked is about Technologies to watch out for right there are new technologies coming in and you mentioned redis specifically as an example uh there is a new database I think vkey uh being very popular now and there are uh so many uh caching systems like dice DB for example is a new open source database that is getting popular uh mcash has been there from a while so technologies will always keep coming the most important thing that I look out for are two things so one is Community Support which is if I'm ading a technology how much of a support I get from Community if I'm facing any issues is there continuous development happening on upgrading it and ensuring that it is secure that is number one and most important thing I look out for and number two is basically the trade of of analysis from performance to cost while I can go for a database that is slightly faster than redis but it might be more expensive for me to actually replace it because I already have thousands of lines of code where redis is already integrated and for my use case red is sufficient I don't need a little bit more faster uh database to give you an example you bought a car you have an engine and you have to replace it with another engine which is slightly faster but for that to do you would actually have to take the entire car apart and then fit it again which doesn't make any sense right it's not equal to replacing your tires or changing the seats because engine is a core component of the car so similarly databases and technologies that to use if they are core components of your system replacing them after a certain period of time is very hard so the suggestion there is to perform trade-off analysis and take a conscious decision sometimes not getting into the hype and following the traditional method might actually be okay because not many people excuse me not many people are actually good at F tuning systems you can actually extract the most performance out of the existing systems or databases that we have we don't need something fancy or faster uh unless we have extracted the maximum value and there is nothing else left and we still more need more that's when you move on yeah thank you it there is a question in chat box that says that a you have said that it had happened 500 million loss in automatic trading system could you explain why it happened is it users money or the company's money yeah so in this case this has happened in uh kns Capital uh few years ago in us that is actually it's a Capital fund management company so it is a it's a bit of both there is a fund that belongs to Knight's capital and there are also investors who would have invested into night's Capital so both their money got lost into the market it not only got lost into the market it actually sort of crashed the market as well because when you see huge amount of buy or sell orders coming in it leads to Market manipulation that was a time in us where there were not tight regulations on the stock market so they had to bring in new regulations to avoid this for example in India this would never happen because in India if uh The Exchange identifies that there is a huge amount of orders which are manipulating the market they would automatically shut down the market or stop trading into those instruments to prevent uh any further Market manipulation or money loss these are baked into our regulations how this kind of error happens is basically since it's a software a buck can actually do anything so in this case there was a bug that got pushed into the production system which actually acted irrationally towards the market and then started placing those orders more questions guys anyone okay so I will I will be moving with the questions that I have uh what are the some common technical challenges while building scalable and secure fintech platforms since we were already discussing about you know how a lot of fintech startups are coming since a lot of ftech startups are coming so the possibility of scams happening is you know kind of increasing so how you will make sure that these challenges we technical challenges we tackle them and how do you ensure security and compilance in an industry that handle sensitive data yeah so India is a fairly regulated uh country when it comes to banking or lending or even trading right so as long as you are compliant with the regulations that are set by the governing bodies like CB or RBI you're okay if you break those compliances you are uh you lose your license to operate so that's how the Indian market works now when it comes to uh the security aspects I mentioned before if you are a fch you need to have minimum three layers of security and uh one of the questions previously was asked right like uh why UPI penetration has sort of a penetration QR based payments have not happened in other countri so one of the reasons is uh QR codes can also be SC scammed right for example you can go to a merchant and replace their QR code with your own personal QR code they might not know if they don't have an audio box present to notify them whenever a payment happens so and a merchant who is not educated might not know the importance of changing the the QR scheme so these sort of scams are very popular which is why uh excuse me so which is why uh Q adoption is not so popular in other countries and in India there are a lot of Innovations being done to prevent such scams um for example when a bank account detects a huge influx of money coming from other people which are not uh which may not sound regular then they would actually flag an alert now how all these systems work is through analysis in the presentation I mentioned that we have both o TP and oil AP databases right in oil TP databases you're storing data in AP data analyzing the data in analysis this is what happens so there is a lot of accounting and reculation work done to ensure that every single transaction is verified if a fraud is detected in any of these transactions then this would be flagged with the particular Merchant and then that particular Merchant or the user will be blacklisted so this analysis process is fairly complex which involves a lot of rules lot of data processing a lot of big data processing that will be involved here and this has to be done by every single player in fch if you're not reconciling your data if you're not analyzing your data for frauds or if you're not detecting any issues then you would eventually lose your license because the government will easily find that you're not really looking after your uh systems and then they'll shut you down you would have seen so many examples of companies which would have started uh popularly and then they would suddenly shut down because something would have happened especially in crypto it is very popular uh you would see so many crypto companies shut down primarily because they are set up as uh fintech companies but uh when you look at the underlying uh assets they would not have reconcile them or process them and so many scams would happen and eventually government has to step in and then disk those platforms from the market awesome awesome so since you belong to that fintech field there was a question that popped up in Q&A uh section that says how secure is this t and pay type of payments what is your thought on that yeah I think uh Tap and Pay uh as with any particular payment system which is physical in nature where uh you would have to uh have a device both users have a device right you need a card which is sort of a device which includes NFC key and then there's also another receiver which is to be needed when there are two physical devices present there is always some of risk present there now uh in case of car payments especially with tap and pay one good thing especially in India is we have certain limits on how much you can pay through via tab for example I think 10,000 rupees is a standard limit that everyone starts with so the worst case scenario are going to lose this 10,000 rupees unless you have increased that particular limit I think most banks restricted to 10,000 rupees now is it really secure maybe not because if you are in a crowded place if you are in a Metro and then if your wallet is not NFC proof then there a chance that someone can basically tap and then capture your information or take in certain payment on your behalf but the good thing with Tap and Pay sort of a system SE the recovery process is fairly fast because you can always raise a dispute and within a span of 30 days RBI would identify the merchant and then recover your money most likely so every payment player who is enabling a tap and pay system should always be responsible for ensuring that the merchant who's capturing this payment is verified if razor pay had a merchant who scanning people Tap and Pay razor pay would have to pay money to the customer because razor pay enabled that payment to happen so at the end of the day if not the person who scamming you the payment player who enabled that payment would also be liable for it and they need to ensure that these sort of merchants are not onboarded into the platform awesome there was one another question now now we are deviating the topic we are going to the finance section but this will be the last question that we'll be taking from the finance section uh so the question was uh how important it is to like why the government of India made it so difficult to invest in US markets so this is definitely not Tech oriented questions but we will like to see your views on it h so it's a interesting already it's a loaded question I would say um it's not actually difficult if you want a certain bit of diversity there are bunch of ways you can actually do it so one way is you can invest in an ETF which tracks uh US Stocks moall ETF I think B 100 is one of those ETFs you can use to uh diversify your assets if you want um but the largely the concern comes with compliance right if you want to trade in US market you need to have you need to be able to trade in US Dollars and you need to also be able to have a US bank account most products that you would have seen which enable you to invest in us have custodial uh framework so what happens is you actually put your money in the uh into custody into the product and then the product will have a US bank account and they would actually buy the US Stock so you don't own the stock which is a huge risk for you because if the company shuts down you don't have access to your stock which is not the case with an Indian stock right if you purchase a stock through grow and if grow shs down you don't have to worry because your actual money uh your actual stock is doesn't in either cdsl or NSL Dem account you can move it to another account and it's fine you not responsible for the upkeep and the availability of the company but in India uh so many sort of uh players started offering US Stocks uh making it look like uh you own the stocks which is not good because uh you don't actually own it you you're giving the custody to someone else there is another question in chat box but again it's from Finance background I think uh let's take it uh let's take it yeah about which you will prefer more crypto Forex Trading or Indian markets like Nifty and why um contrary to what people think I'm actually pretty least interested in uh investing or trading or stock markets one of the reason reasons I started working at Market field is because I don't like spending a lot of money understanding learning market and trading I like Building Products I like shipping code I like um technology I don't like Finance as much which is why it makes more sense for me to work in ftech because I make others like me happy by ensuring that they don't have to worry about all of this so in my in this case I don't prefer crypto or Forex or trading in any Market I don't trade at all uh I've not traded once in my life uh most of the trades that have happened in my account are through automated trading because uh it's a hugely emotional process whether we like it or not and it requires a lot of time and effort so um yeah uh and I prefer uh working with experts so there are a lot of financial experts uh there are research uh analysts who are present who are registered by uh C and there are a lot of portfolio management companies out there who have experts who studied these markets and they're good at this job so I'd rather my money be operated uh and kept by people who know more about managing money than I do uh and I want to do what I'm good at which is coding that was quite a contradiction so uh this before ending the session this will be the last question that I'll will be taking the question is uh for students new new to both finance and Tech do you have any uh suggestions like projects that we can develop and we can show show it in our resume for for someone like me who is very interested in finance so so that we can represent our skills of both finance and Tech since you are from that background so can you suggest a simple fintech project idea yeah uh yeah I think uh so again fintech is also a broad space right it includes payments uh payou

2025-01-26 07:48

Show Video

Other news

You ever seen Cat OS? Connecting to the Internet in 2025 2025-02-14 04:30
This $4.4bn Technology Is CHANGING Aviation 2025-02-12 19:41
How Infrastructure is Powering the Age of AI 2025-02-09 20:43