● Worked on a high-throughput training system that optimizes efficiency and resource utilization of Reinforcement Learning algorithms by up to 20% across games and simulations than the state-of-the-art systems (accepted at ICML 2020).● Research was based on applying on-policy Reinforcement Learning algorithms to control a swarm of Quadrotors.● Built a faster alternative to Python multiprocessing queue and also worked on CI pipelines for ML Research Projects.● Worked in tandem with the Autonomous Vehicle Lab at USC on Navigation and Path Planning for an autonomous car to participate in IGVC 2020.

● Worked on designing and developing a platform that makes it easy to get high-quality training data for Machine Learning models at scale.● Researched on integrating Machine Learning to automate certain Q&A capabilities.

● Optimizing and writing SQL Queries to fetch data for various internal analytical purposes.● Using Cognos And Salesforce to generate Business Intelligence reports enhancing Fundraising and Donor campaigns.

● Worked as a part of the Steel Security Research Lab on an application called DEW (Distributed Experiment Workflow) that aids information security researchers in running experiments on multiple testbeds via a Distributed setup.

● Designed and developed full-stack architectures for more than 5+ products as a part of a two-member team using Python, Angular, and MongoDB.● Spearheaded the development of real-time financial backtesting frameworks using Redis and WebSockets for trade systems.● Created and deployed trading strategies for the Indian Equity derivatives market that had consistent daily returns of 2-3% and also for the US equities market on the CloudQuant platform that received an initial capital of $5 million.● Created ML-based trading strategies using XGBoost and Outlier Detection models with research focusing on developing strategies using Multi-Agent Reinforcement Learning with Policy Gradients.● Built a proof of concept to increase the fuel efficiency of garbage trucks by 15% for a client using Combinatorial Optimization and Reinforcement Learning based on Constrained Policy Optimization.● Worked on Drone-based Image Processing solution for road quality detection and traffic counting using Deep Learning with the YOLO algorithm which achieved an accuracy of 84%.● Built a Computer Vision-based OCR system for document processing which had a Generative Adversarial Network(GAN) for preprocessing and achieved an accuracy of 80%.

● Optimized the web application of the Coverage & Investment Risk team by building a new loan processing framework using Java and Angular that reduced processing time by 30%.● Piloted the development of a chatbot framework with the architecture team which reduced the number of queries to the HR team by 25%.● Developed a Deep Learning-based text summarization proof-of-concept using LSTM which generated reports of companies' performance to reduce the workload of analysts and saved $25 million.

● Developed an application that automatically detects inconsistencies in SI unit names of product design documents.