Hello, I'm Brij Mukesh Mangukiya

SmartTechPrep (work in progress)

An AI platform to practice technical interviews including leetcode style questions. AI agents to evaluate your response, check code, ask follow-ups, determine time and space complexity and give out scores. React-based frontend with in-browser IDE. Python backend with use of Langchain agents, FastAPI middleware and SQLAlchemy SDK to store data on Supabase PostgreSQL. Hosted on Netlify and Google Cloud Run, with ability to run locally using Ollama

AirBNB RAG Recommender

A web application that takes user queries and recommends Airbnb listings based on similarity search. Utilizes Large Language Models to select the most suitable listing within the top-k similarities. Uses FastAPI and Jinga templates for UI. Numpy for calculating cosine similarity. Locally hosted embedding(BAAI-bge-large-en-v1-5) and LLM model(Meta-Llama-3-8B-Instruct-GGUF). This entire project is hosted locally on a home-server with NGINX, so availability may vary.

GenTerm

A terminal-based RAG Q&A platform written in Go and JS, that allows users to upload documents, images, and other file types, and ask questions about them in a terminal interface. Hosted on Azure Web App container.

OpenResumeAI

LLM API middleware

Written in GoLang. Enables secure use of locally deployed LLM and image-gen models over the public internet. I wanted to make use of my desktop GPU for LLM models when I was not on the same network.

Watch-List web app

Created a .NET web application using MS-SQL and TMDB API, enabling media discovery, and YouTube embeds for trailers and snippets. Integrated Google Charts to visualize user preferences for genres and media, leveraging database views and stored procedures for dynamic insights.