Oct 2017 – Mar 2023 · 5 yrs 5 mos · Bengaluru, Karnataka, India · Hybrid

•  Developed stacked ensemble classification algorithms to derive the operational factors influencing faulty packaging of a delta robot in the supply chain factory. IoT data produced by sensors strategically located on machines. The data was collected under ‘normal’ as well as ‘abnormal’ operating conditions. This helped reduce the faults in the robotic arms while packaging by 15% and therefore save re-packaging costs.
•  Researched about the Neo4j Graph Data Science (GDS) library that contains many graph algorithms like similarity, centrality, and link prediction. Built a product recommendation system using FastRp and kNN algorithms. For each pair of similar customers we then recommended products that have been purchased by one of the customers but not the other, using a simple cypher query.
•  Developed Sentiment Analysis using NLP to analyze comments given by IPO R&D applications users. A monthly report has details like Ticket Backlog, Ageing Tickets, In Time Resolution, L1 Fixed Issues, Reopened Issues, Customer Survey Report and Comments. Applied NLP technique to the comments given by end users to determine whether sentiment is positive, negative or neutral text classification.
•  Built, trained, and deployed machine learning models at scale. Deployed ML model in AWS SageMaker and other services like AWS CodeCommit, AWS CodeDeploy, AWS CodeBuild, S3 and a variety of tools and features to help us manage and monitor our models, including Jupyter notebooks, data visualization tools, and automatic model tuning.
•  Legacy enterprise apps migrated from costly on-premise servers to AWS EC2, DMS, ECS/ECR, S3, and RDS that resulted in decommissioning costly high maintenance on-premise servers. These on-premise servers were costing 250K USD pa just for maintenance cost but now with AWS automation, it is Serverless/On Demand and costs us about 60K USD pa.

Jun 2012 – Sep 2017 · 5 yrs 3 mos · Gurugram, Haryana, India · On-site

• Trading Analytics & Research is a technology focused team that does in-house research Blackrock's on own order and execution details from proprietary Aladdin application. This helped in assessing portfolio performance and predicting trading performance against expected costs.
•  Modelled stock portfolio performance with Python. Built Time Series analysis to see how to better evaluate the financial performance of customer’s investment portfolio against the stock market index such as S&P 500.
•  Calculated the Capital Asset Pricing Model (CAPM) with Python. We first defined several key variables that we need for the CAPM formula including the risk-free rate of return, the market return, and beta. We then looked at how to calculate the daily returns for each stock, calculate the beta of an individual stock, and apply the CAPM formula. Finally, we looked at how to use the CAPM formula for an entire portfolio of stocks.

Jan 2011 – Feb 2012 · 1 yr 1 mo · Hong Kong SAR · On-site

• Position Services is a business vertical within IT development & operations which manages the firm’s risk around Corporate Action events for Equities and Bonds. Utilized ML techniques for integrating data from diverse sources, including market feeds, news and historical data, to create comprehensive datasets for analysis. Implemented anomaly detection algorithms to identify unusual patterns or behaviors in market data around the time of Corporate Action events. This helped in early detection of irregularities or potential risks. Developed predictive models to forecast the impact of different Corporate Action events on stock and bond prices. Applied time series analysis to model and forecast time-dependent trends in market data, especially around the occurrence of Corporate Action events. This provided insights into the temporal dynamics of market behavior.

Jan 2009 – Jan 2011 · 2 yrs · Singapore · On-site

• CASTLE stands for Capital Adequacy, Statistical-reporting, and Large Exposures. It’s an application used to calculate the bank' risk exposure to regulatory norms of different markets. Developed ML models to assess credit risk by analyzing historical data on loan defaults, credit ratings, and other relevant factors. Utilized classification models to predict the likelihood of default and estimate the potential loss in case of default. Used time series analysis and risk factor modeling to capture market dynamics and fluctuations. Implemented natural language processing (NLP) for extracting relevant information from regulatory texts. Implemented real-time monitoring systems using ML to track large exposures dynamically and provide timely alerts for potential breaches.

May 2007 – Dec 2008 · 1 yr 7 mos · Bengaluru, Karnataka, India · On-site

• CCT is a confirm generation and tracking solution that allows automatic generation of trade confirms for transmission to various counter-parties. Used OCR to extract information from scanned or image-based trade documents. Combined OCR with NLP to convert extracted text into structured data for confirmation generation. Implemented named entity recognition and information extraction to identify key details such as trade date, instrument, counterparties, and terms. Employed text classification models to categorize incoming messages into different types (e.g., trade confirmations, amendments, cancellations) for appropriate processing. Utilized anomaly detection models to identify unusual patterns or outliers in trade confirmations that may indicate errors or fraudulent activities.

Jun 2005 – Apr 2007 · 1 yr 10 mos · Bengaluru, Karnataka, India · On-site

• Summit is a trading and processing solution that allows users to consolidate businesses, minimize operation risk, reduce transaction processing costs and obtain intra-day risk management information at the desk or enterprise level. Analyzed market data to identify trends and insights that can inform trading strategies. I implemented data warehousing and ETL (Extract, Transform, Load) processes for seamless data consolidation. Defined and tracked key performance indicators (KPIs) to measure the success of business consolidation efforts. Define and monitor data quality metrics to ensure the accuracy and reliability of data within Summit. Monitored transaction volumes over time to identify patterns and potential anomalies. Implemented interactive dashboards and reports for users to visualize key metrics and analytics results.