My Projects
This page highlights a few projects where I’ve focused on backend engineering and GenAI / RAG systems—from AI chat platforms to document and web‑based assistants, and a production‑style academic portal.
Each entry summarises the problem, what I actually built, and the engineering decisions that mattered, so you can quickly understand how I think about system design and implementation.
Lumin – AI Chat Assistant
Problem: One‑off chat demos don’t scale when you need multiple tools,
models and workflows behind a single interface.
What I built: A full‑stack chat platform with a React
frontend and FastAPI backend that routes user messages through a modular
LangChain/LangGraph graph, with MongoDB for session and message history.
Engineering decisions: Clear separation between UI, API, and tool layer;
typed FastAPI routes; and a pluggable tool registry so new tools/flows can be added without
rewriting the core application.
WebQuery – RAG Browser Assistant
Problem: Reading long documentation pages and FAQs is slow when you only
need specific answers.
What I built: A Manifest v3 browser extension that lets
users chat on top of any supported documentation page, backed by a FastAPI
RAG service with Qdrant for vector storage and BeautifulSoup for scraping.
Engineering decisions: Per‑site collections in Qdrant, chunking tuned to
documentation structure, and a thin background script that keeps all heavy work on the
backend API.
PDFQuery – Ask Questions
Problem: Searching through long PDFs manually is slow and error‑prone.
What I built: A Streamlit frontend and
FastAPI backend that let users upload PDFs, index them into
Qdrant, and query them using a RAG pipeline, with answers streamed back
to the UI.
Engineering decisions: Containerised Qdrant via Docker,
separate services for UI and API, and simple configuration to swap between different LLM
providers without changing the retrieval layer.
Course Registration System – IIIT Vadodara
Problem: Needed a system that captures the real course registration
workflow used at IIIT Vadodara, with different roles and approval steps.
What I built: A Django web application with
role‑based dashboards for students and faculty, session‑based authentication, and Gmail
notifications for registration actions.
Engineering decisions: Used Django’s auth and permission system for
roles, separated student/faculty views, and designed the models around the institute’s
real processes rather than a toy example.
Ideal Project Management Process
Planning & Research
I begin by understanding the project requirements, conducting research, and planning the architecture and design.
Design
I create wireframes and mockups to visualize the user interface and experience before implementation.
Development
I write clean, efficient code following best practices and implement the planned features.
Testing & Refinement
I thoroughly test the application, fix bugs, and refine the user experience based on feedback.
Deployment
I deploy the application to a hosting platform and ensure it runs smoothly in a production environment.