full-stack AI engineer.
I build software at the spot where AI stops being a demo and starts pulling its weight — agentic backends, OCR pipelines that catch real money, desktop apps shipped as one-click executables. Mostly small, ambitious things; the kind that take longer than they should and feel cheap once they finally work.
- Now shipping
- PMO Insights @ Accenture
- Latest side-quest
- Claude-Pro-as-HTTP-API
- Thinking about
- MCP, agentic SQL, sub-second LLMs
- Operating from
- Bangalore, IST
A quick hello.
I write code that lives close to users and decisions. Most of what I build ends up as a Chrome extension, a desktop app, or a small API on a free VM — specific enough to be useful, small enough that one person can hold the whole shape in their head.
Right now that's PMO Insights at Accenture: a FastAPI + Electron platform with a Claude-via-MCP brain that PMO leadership actually uses to find risk in their workforce data. Before that, BARCGPT — an OCR + LLM pipeline that caught ₹2L+ in faulty reimbursement claims by reading scanned medical bills better than the humans were.
I'm biased toward the pragmatic tool over the popular one. MCP over RAG when the data is structured. PyInstaller when end-users don't have Python. SQLite when Postgres would just be theatre. I demo earlier than feels comfortable, push back with a working prototype rather than an argument, and try to keep the JSON parseable when the LLM doesn't cooperate.
Off-hours: I'm probably writing a small browser extension to fix something that annoys me, compressing a WhatsApp file under 2MB without leaving the chat, or solving Tatkal captchas with OCR because IRCTC refuses to be reasonable.
Things I've shipped.
The tools
I reach for.
Python and TypeScript are my daily drivers. C#/.NET from the Accenture training stack. Everything else is picked per problem.
A short history.
- Oct 2025 — Present
Advanced Application Engineer (AEH) — Accenture — Industry X DS, Bangalore
Full Stack AI Developer. Sole developer on PMO Insights — gathered requirements from senior PMO leadership, shipped a FastAPI + Electron desktop platform with an MCP-backed Claude chat layer. Currently in UAT.
- Jan 2025 — Jun 2025
AI/ML Research Intern — Bhabha Atomic Research Centre (BARC), Mumbai
Built BARCGPT — an OCR + LLM document intelligence system. Integrated FEATUP into PaddleOCR for 1.5× readability, ran Mistral-7B in 4-bit on an 8GB GPU, and pushed inference latency from 10s to 1.5s with DSPy.
- Jun 2024 — Jul 2024
Data Analyst Intern — Bhabha Atomic Research Centre (BARC), Mumbai
Analyzed 10,000+ housing records under DAE; designed and shipped 8 Grafana dashboards backed by CTE-heavy MariaDB queries for live occupancy and allocation insights.
- Aug 2023 — Oct 2023
AI/ML Intern — Telekinetics, Pune
Built a Python resume-parsing system processing 45,000+ candidates with 95% accuracy via Flask + spaCy. Identified 20+ pre-launch bugs on Turant, a voice-authentication app before its global release.
- Jun 2023 — Jul 2023
Infosec Intern — Nuclear Power Corporation of India (NPCIL), Mumbai
Deployed and configured Snort (open-source IDS) to monitor and triage network threats with the incident-response team; grounded the work in firewall and access-control fundamentals.
- Jul 2021 — Jul 2025
B.Tech, Computer Science — VIT Vellore — CGPA 8.64 / 10
Selected to Amazon ML Summer School 2024 (3,000 from 60,000+ applicants).
Let's build
something good.
Best place to reach me is email. I read everything and usually reply within a day or two.