Presently engaged in the AzureDevops Copilot project, where the primary objective is to completley eliminate the human effort in PR (Pull Request) reviews by analyzing data signals from Pull Requests, Build Logs, Test Suites and about 8 other data signals. The goal is to explore the potential integration of these successful features into internal Microsoft repositories hosted on Azure Devops. Using technologies such Kusto, C#, Bing trained custom LLM Model, Azure AI Search. Prior to this project, I worked on features to allow the users to create PR summary of Github, generate test cases, and expose vulnerabilities in the code using Prompt Engineering, Azure OpenAI APIs, and LLMs. Worked with the Playfab team to enhance user retention by using ML algorithms such as LightGBM classifier. The idea is to allow the Game developers to predict the player churn probability so that our customers such as Unity, and Roblox have the opportunity to keep the player engaged. The customers confirmed that ML solution helps them retain 80% of the players who had a moderate to high probability to churn. Worked on pixel streaming technology to allow users to not only stream games just like a video on youtube but also interact with the game using a keyboard and mouse. Worked on features such as adding a layer to support Virtual networks so that the private preview games/videos of major clients such as Disney could be tested locally, making the system fault-tolerant with supporting fail-fast deployment, and dashboard for the clients so that they could change the certain settings such as scale threshold, frames per second while giving a seamless experience to their customers. using python and PowerShell. https://docs.unrealengine.com/5.1/en-US/pixel-streaming-in-unreal-engine/

Worked in Ignite/Soju team. I helped deliver and productize the Detail Page for the AWS courses launch and buy box-related experiments. I worked with NLP team to store user feedback fetched from the course view portal. I also worked on Amazon Ads and Course preview features to help the course authors to sell more courses.

Worked on Data Analytics project to visualize and analyze the traffic profile of various services under office org to free up the unused Azure Capacity for external customers to meet the additional requirement of customers during the first wave of Covid. Worked on scaling and containerizing PowerPoint build infra using Kubernetes and Docker

Built async workflows using C# and Azure services to handle the Neural Net Training. Built translation models for Custom Translator, a product that allows user to train the Neural Net given the custom data set. Responsible to handle production bugs, training related issues and directly answer user's query.Extensively worked on making the async training workflows fault-tolerant on Cloud, bring down training error percentage from 60% to 5%

Worked in Azure Machine Learning Services team to build tools for data scientists so as to make ML model scalable by deploying them on Azure, monitor test data set on production to give data drift insights of the performance of the model using Apache Spark and python

CSCI 353(Introduction to Networking):Held discussion and doubt sessions. Responsible to come up with course material/syllabus. Also designed coding assignments to be given to students and automated testing scripts.CSCI 310(Software Engineering): Responsible to hold discussion for students to discuss software requirements for a specific project with students. Also responsible to discuss pros and cons of specific software architecture/technology used.

Worked with Autonomous Infrastructure Management(AIM) team to design an event/pub-sub based service using Python, Apache Thrift and facebook's internal tools to automate and coordinate various stages of server maintenance. Deployment of this service on 10 racks in North Carolina data center enhanced CPU utilization by 1%.

Responsible to code, maintain, test, debug APIs for the App Welconnect. Enhanced the Logging Architecture of the backend services. Implemented Feedback API to make the app more interactive. Optimised SLA of the various APIs on an average by 60%. Also I helped the team to normalize the database tables in Postgres and hosted the same on Heroku.

Held discussion and doubt sessions CSCI 353(Introduction to Networking) to undergraduate student. Responsible to come up with course material/syllabus. Also came up with coding assignments to be given to students and automated testing scripts

Worked on Lodging team as a Java Developer. Developed Java REST APIs to access hotel data using Spring Boot and Jersey to help hoteliers and market managers of large hotel chains to manage their respective booking data. Later moved on to revamp the entire hotel booking tech stack of Expedia and Affiliated Networks from C++ to Java.Implemented logging architecture using ElasticSearch, Kibana and Logstash.

SPEAR (Structured Products Evaluation and Research) is a suite of applications that is used byBlackRock's internal and external clients for modelling of mortgage backed securities.Implemented desktop based application using MVVM and WPF architecture to analyze commercial mortgages. Responsible for gathering requirements,designing, development, unit testing and monitoring overall health of the system