Zepto Policy RAG Service
A retrieval service that answers policy questions in JSON a caller can trust.
- LangGraph
- ChromaDB
- sentence-transformers
- Pydantic
- FastAPI
- Docker
Hi I’m
AI/ML • Python • Full Stack (MERN) • GenAI, RAG & LLM Agents
B.Tech CSE (AI & ML) student building retrieval systems, LLM agents and the full-stack products around them.
Elsewhere
I build retrieval systems and the applications that wrap around them. Most of my time goes to two things: getting an LLM to answer from a source of truth instead of improvising, and shipping the interface that makes that answer useful to someone.
That has meant a LangGraph pipeline over ChromaDB that returns Pydantic-validated JSON so downstream code can trust it, a real-time pose-estimation coach running MediaPipe in the browser loop, and an arbitrary-style transfer model trained on Kaggle when I ran out of local GPU. On the web side I work in the MERN stack and Next.js — auth, protected routes, WebRTC, the parts that break in production.
I'm heading toward graduation in 2027 and looking for work on production ML systems — evaluation and retrieval quality specifically, where the interesting problem is usually measurement rather than modelling.
—
B.Tech, Computer Science & Engineering
Artificial Intelligence & Machine Learning
CGPA 8.15 · Hyderabad, India
—
Intermediate (MPC)
95.8% · Hyderabad, India
—
Secondary School (Class X)
CGPA 10 · Hyderabad, India
The tools I actually reach for, grouped by what they're for.
Newest first. Hover or open a card for the longer version.
A retrieval service that answers policy questions in JSON a caller can trust.
A real-time gym coach that watches your form and corrects it mid-rep.
Arbitrary multi-style transfer in a single forward pass.
A Hacker News pipeline that fetches, categorises and tracks story trends.
Peer-to-peer video calling over WebRTC with a signalling layer of its own.
A rental listings platform with maps, image hosting and reviews.
DeepLearning.AI & Stanford Online — Andrew Ng
NPTEL
Microsoft
NPTEL
NPTEL
Open to internships and new-grad roles from 2027, and to interesting problems before then.