ProApplyAI is a web application that uses AI to write personalized cover letters. The webapp simplifies the cover letter writing process to merely pasting the job details The technology stack includes NextJS, Redux, and TailwindCSS for the frontend, and FastAPI, Langchain, and ReportLab for the backend. MongoDB is the primary database and GCS is used for document storage. The frontend is hosted on Vercel and the backend on Google Cloud Run. NextIntl is used to support multiple languages, Stripe is used for payment processing, Mailgun for email notifications and the website is fully responsive.

The Fault Tolerant and Scalable Surfstore project combines the Raft consensus algorithm for fault tolerance, availability, and consistency of the metadata service, along with consistent hashing on multiple blockstores for efficient load balancing and distribution of data in a distributed storage system.

SurfStore is a blob cloud storage distrubted system service fully written in Go using gRPC. (Similar to Drop Box). It uses MetaStore to store metadata and a BlockStore to store blob data.

MERN stack (Mongo Express React Node) website that tracks Cryptocurrency prices and news. The site allows users to sign in and customize their dashboard of crypto currencies and has authenticated routes with JSON Web Tokens. It also keeps users logged in with session cookies. And stores passwords safely using bcrypt.

Created a website that summarizes text using Natural Language Processing (NLP) with Django, SpaCy, and NextJS. Hosted on AWS Elastic Beanstalk and Vercel.

With a team of 4, created an IoT Smart Iced Coffee Machine MVP product. Developed hardware and IoT features, and presented our business plan that won us best startup team project in a class of 200 students.

Implemented an efficient 2-D particle filter SLAM from scratch using lidar, wheel encoder and gyroscope data from a self-driving car. Also implemented texture mapping collected from stereo cameras. Wrote a paper describing the implementation.

Implemented an efficient Extended Kalman Filter (EKF) SLAM from scratch using IMU data, and Stereo Camera data from a self-driving car.

With a group of 4, recreated the famous 1979 arcade game Galaxian to be played in the terminal. Fully developed using functional programming language Haskell and the Brick library.

Implemented asynchronous multi-agent parallel reinforcement learning algorithms: Q-lambda and Q(0)-learning using C/C++ and CUDA.

With a teammate, used techniques (convolution via FPGA verilog, GPU parralelization CUDA) to accelerate ML algorithms for forgery detection in media with hardware.