With the release of many extremely good large language models across the internet, I decided to build a chat application that lets you choose the model you want without paying for many different subscription. With this app you can add any OpenAI API compatible models that you bring on your own with API keys and chat with it seamlessly with a great and responsive mobile friendly interface with PWA support. It also supports upto 5 team members to brainstorm together with a LLM in a Conversation. There are many optimizations and best practices I learnt from buliding this app and i plan to keep adding more features to it and turn it into a Microservices architecture going ahead with a backend in Go and LLM stuff in Python. It has more than 140 users.

Developed an intelligent terminal based stock analysis tool that leverages advanced LLMs with tool calling to dynamically retrieve real-time financial data via the yfinance API.Integrated LangChain with Gemini 2.0 flash to enable natural language queries and deliver insightful market analysis across 125 different stock related values. Used plotly graph objects to visualize multiple stocks in a given time frame in a interactive plot

Developed a Twitter clone to test my UI development skills as well as learn a different way to communicate with backend systems using GraphQL. The project was built using Next.js for the frontend and Node.js for the backend. Prisma ORM was used to interact with the PostgreSQL database. The project also utilized Redis for caching and rate limiting, Google OAuth for authentication. The project was deployed on AWS in a scalable manner. The AWS instance may not be online as there are no users to use the app.

PDFQuery is a web application that lets you provide context to the LLMs from PDFs using RAG The application uses LangChain for language processing and FAISS for similarity search. The application is built using Streamlit and Python. The application also uses gemini-flash-2.0 model in the Gemini API for the LLM.