..

Projects

website chatbot

As part of moveworks organized hackathon, we built a chatbot that answers customer queries based on information on the entire website.

Used everything that is open source / can be run locally: sentence transformers to generate embeddings + mistral for LLM output running locally on a laptop + similarity as well as tokenized search all wrapped as a flask app. github

implementing transformer

Implementing the paper “Attention is all you need” in scratch using pytorch. github

finance gpt

Ask it anything regarding finances of a publicly traded company and it will answer!

A simple flask app that uses OpenAI embeddings and faiss DB to store vector embeddings to then query and answer user query. Made it under 30 hours as part of a hackathon. github

simple note taking app

Built a simple note taking app to get famliar with MERN stack in 1-2 days. github deployed here

gpt answers

I did not want to keep on answering my younger brother’s science and language doubts so made a simple website based on gpt-3.5-turbo api so he can ask and get direct answers easily. You can check it out here: solve.vimarsh.co

fee ninja

A crypto analytics platform that tells you how much you have spent on gas fees on different blockchains. link

A full stack app, that searches through transactions for your wallet and returns how much you have spent on gas. An express (self-hosted) backend and vanilla HTML, CSS, JS based frontend.

plantogen

PWA that can detect plant diseases (completely offline and on-device) link

hardware stuff