Hello everyone. I'm Sona. I have working as a data solution lead and a data scientist, so I would like to discuss about a meets ml. How the transforming financial decisions by using neural network to what is the financial decision intelligent as it is used for risk assessment. For example, credit scoring and fraud detection and investment strategies. Some algorithms are trend tradings, et cetera. And we can forecast market trends and operational decisions, how the loan approvals can be given. So by incorporating this A ML with the neural network, so it fast the decisions and pattern recognitions can be also happen. And ha it handles noisy and incomplete data. And accurate predictions in the real times. So we will go into deeper the topic how we have been implemented, and what can be the chances of results. About myself. So as I mentioned, I have been working as a data scientist and mostly in the banking and finance two minds. Whereas I had gone through multiple use cases and the solutions given and which is going very well, and live the projects also has been happen with the solutions. What proposed? What is the role of a neural network? This neural network is it's a human brain. It's this inspired with the neurons. It thinks as like a human brain. These are the capabilities. It can handle huge data sets, and it captures non-linear relationships also. And definitely it's a real time how the human brain can think about realtime processing the same thing. And there are some types of neural networks. There's RN. It's mostly, it's for time series predictions and CNNs are basically, it's for the chart patterns and the transfer models. It's both sequential and complex data relationship and it handles internally when we incorporated with these decisions. So it'll take a neural network into the considerations when we choose an option for it. And I would like to talk. Talk about some of the transformations the banking domain, how we, how it has been getting used of this. For example, this aid driven risk assessment. So how this past processing the data and how it has been banks and financial companies. Uua, aI to scan through huge amounts of data, right? So it's internally how the data is instantly the spots they can identify how the customers can be defaulted. So these kind of risk assessment can be detectable and anomaly detection. That is something which we can easily identify what is the false I. Alarms. And though it is showing as a good, but there might be internal is some algorithms which will trigger yeah, the positive positivity of anomaly detections. So investment strategies also AI can trend and learn from the past rates and consistently improve the strategies like, trial and error. It is nothing. But this helps investors and financial firms make better decisions about the data and when they invest money sometimes even pitching human traders. Executive intelligence is strategic decision cycles. This helps company leaders make smarter decisions fast and faster. This AI can predict trends, market changes and customer behaviors. And so executive don't need to wait till weeks for analysis. They can act with the real time data and for intelligence. So these are the major transformations we can consider in the bank. And the risk assessment engines. So if we go for this pyramid, so if I go for the real data time processing, this is considered structured and unstructured structured can be defined is a better way since it is A-T-P-M-S and very structured. So those can be handled, but the unstructured and semi-structured data is quite challenging. We, with the standalone systems, it's hard to do this. With this cloud and AI solutions it goes with the smooth processing and AI created transfer learnings is something, whatever the knowledge it is take what if, what it learned from the one area and it applies to another area. If, for example somewhere the products happened and it mentioned this kind of scenarios is happening, so you may get a. Chances of this default, so that kind of previous it have a repository in that and it'll come into the f in other location. It happened the same type of scenario. It mentioned that, okay, if this is leading to your situation also. So this kind of guided transfer learnings also can be happened and advanced prediction capabilities. If you go for the system doesn't just detect what's happening, then now it predicts the later dates, how it can be happen, and what is the solutions for it, these extra futures it predicts and give it help. So in summary, this handles any kind of. And it's learned from different domains, not only from single bank domain, but it consider all the domains and this predict future problems and keep improving over the time. It's not a one time results and there are some components in a ML risk. So dynamic risk scoring is something we can update risk profiles that adapt some threats. For example, this means that the AI systems keep updating a person's or company risk level as new data comes in. So it, doesn't just use the whole credit scores or reports. It constantly check new patterns. And it applies. So for example, if a customer who suddenly starts spending a lot of more money and few payment might be flagged as a higher risk. So that brings to the situation and scenario simulations is this where the system imagine many what if the situations to see how bad things could happen, and it tells banks to plan and advance more safety measure. So the AI simulated stock market crash, for example, are rising interest rates are major economic slowdown. F then checks the bank's investment would perform each case. I'll pick the bank plan and smarter strategies also and explainable outputs. Sir, this is not something, the black box, it happens. It's giving the results. No, it takes explain the detailed information. For example, if the customer is denied for a low, then it mention why it has been denied. This income is low. Or it's too many recent credit inquiries are late payment for the past six months. All these kind of situations, it explain so that it's a justification. And if I take one of the examples, this is the anomaly detection framework. So what here I can mention that there are behavioral biometrics. So behavioral biometrics is something AI watch how people interact with other they devices and confirm. So if it is accessing if you are usually, ask the question or type quickly, but somewhere you just became very slow or else some, sometimes you are making urgency of something so it identify this behaviorals and if some, someone. You are intense. Something hit the phone shake and when you are talking, so these kind of also its aspect and transactional stress is something as looking for the pattern. How usually. Spending the behavior. We can say that if you usually show up in your city at one local store, suddenly you started spending high amount and uneven times and all. So then it identifies and triggers that, okay this is something is unusual and cross channel activities. So it's a tracking behavioral. Come across the behavior states, same across phone, apps, website, and devices. If you log from multiple devices from one geographical location, for example, us from different locations, different states at a time. So then it suspects why this has been happening. So something is efficient and network analysis. Also, it can be done, it finds the multiple accounts are making a similar purchase. Transferring money to the same group. This might be a fraud ring. If two unrelated customers suddenly send money back and forth multiple times, it might be many hundred, right? So these kind of things it's identified basically in anomaly detections. It's catch pro early stage and reduce false alert. And protect users without slowing down the transactions. AI announced investment strategies is the sum thing. It's a real reinforcement learning. Reinforcement learning. That's the optimal solutions. That strategies, it keeps learning. It's keep learning and explain very well for us, which is the better solutions and where it can be, go for it in a better plan. And it learns from the previous models. It's keep learning and it's a mature. That time and NLP sentiment analysis, it's a text processing, how you type in the text. And is there any common type of words or this, which is giving sensing for something else, pro or something. So just it'll give a sense of from this, your text, whatever you are typing. So personalized pro. Portfolios truly that in whatever the investment recommendations are aligned, this any tolerances are there, or any financial objects behind this model portfolio. This reinforce investment strategies and this algorithm that continuously learn from the market condition and transactions outcome. So by combining all this machine learning with the human achievements and these all deliver the sophisticated strategies and continuously available, only the investment investors are net worth individuals and which is they can get benefit out of this and performance during market, is something from the past three years, the performance has drastically increased. Now every of the financial companies moving from traditional approach to the strategies because instead of spending a lot of time for this analysis, data fetching and individual vendors dealing all these. Making them for long time processing and the, they have to wait for the results for long time too. So now whoever is opting this one, 32% is increased from the three years to expecting more in the upcoming years. And past performance is something that's a response. Time is market shift is a is a fast because every banking transactions are everything. So they wanted to make a decision very quick and at the same time it has to be updated. So that response and eventually the client retentions will be good because if they are satisfied with the service and it's leading to their solutions, then definitely the retentions will be high. So that is what the 65 to 70% aspect research. Execute this intelligent platforms is give a predictive analytics if forecast, what is the business outcome and how it can be happened. So it predicts and give the solutions and scenario planning is something. That you can plan for whether this kind of scenario it fits, and how this modeling which future can be used for this to execute into the real time and decision support is something, what are the strategies which you can take? Place a fit. And every use case is different. And it with respect to the financial domains and that use cases, they need to go for it. And they can take decisions And performance tracking is something that they will have a measurable outcomes. It's, we can see it rapid performance implement by excluding this process and this predictive analysis. Analytics also giving a better decision in the organization levels if you wanted to implement this framework. So first we need to identify the use case, whether it belongs to our organization or our financial, or especially the use case which we are planning because every problem statement will not fit into the A ML. So first we need to under understand what is the purpose of. To move this a ML and your network. So why it has been Because that we need to identify first. Then after that, the in infrastructure preparation. So whether you wanted to go for cloud, if cloud what, what else is a Google cloud? And AWS says you, which platform you wanted to choose it, that you can decide. And most of the things. The A ML all these services are matched with the cloud platform, so we need not to go first separately, so they, we will have one bundle and deployment strategies also very quick with us. And pilot deployment as take samples to and flow, you can develop with some sample set of data and you can see the results. If you satisfied with the small case use cases, then that pilot project can be deployed and monitored for some. Time. It is not something just we have deployed and cannot go for it. I just will monitor it and observe whether the patents are tallying and the expected outcomes are reaching out and scaled. Implementation. You can go for once it is the pilots across successful and you can follow the governance and monitoring and treating affected approaches. And then the, we can implement the approaches in the best way. So the most effective deployments is is a dual focus on technology and people and organization adoption. The skill management development is also required, and we need to have a clear communication of bandwidth or what exactly we are going to get of this technologies. I would like to call out some key takeaways from this session. Say, hybrid is a superior. Definitely. We cannot just simply go with the one one adoption. We can combine AA and ML techniques and neural networks in between. Then underlying one single approach just to, we develop one kind of model so that will not give better results. And even we cannot depend on one. One application and you can have it. Start with good data good data in the sense we will, we have more noisy data and identify what are the quality of data and what are the parameters can be used to take the decisions and whether it is it passing the governance policies, for example, this SSN and all we cannot use directly the masking policies and all will be in place and then we can go for it and invest meant for this. Data infrastructure before we are passing the inputs to the model. And Human Plus mission partnership is an important definitely one. Mission cannot implement and directly go for all the decisions all the time. The women intervention is required. We need to pass it judgment so that it'll learn it from there and give the solutions and measure continuously. It's not something just implemented and done so. No, it's just you need to go for all the time. How the regularly it's a ized feeding and regular patterns. Also, we need to feed it. And complexity scenarios also will have to measure continuously. So this financial institution successfully implemented this combinations, and they got an advantage of this operational efficiency and risk management capabilities. To begin with this a ML journey. Understand the requirement as capability assessments and high value use cases, and create a roadmap and infrastructure and go for it. Success. Thank you for giving this opportunity. Yeah, we, I would like to mention if any of your use cases or solutions is to be addressed, feel free to contact me. Thank you.