Just an old, blank piece of parchment. Nothing to see here.
Tap your wand anywhere on the parchment to reveal it.
(or press Enter)
Mr. Commit wonders what you're waiting for.
Messrs. Code, Coffee & Commit, purveyors of aids to magical developers, are proud to present
Final-year ENC student @ Thapar · builder of firmware, models, agents & the systems that keep teams running
Engineer who also runs the operations: I build the product, the process and the playbook behind it.
Tip: this map answers to wand gestures. Hold the mouse button and draw a circle anywhere.
follow the footprints ↓Room I
I'm a final-year ENC (Electronics and Computer Engineering) student at Thapar Institute, graduating in 2027. I build across the whole stack, from STM32 firmware and PCBs up to machine-learning models, LLM agents and web apps.
I've written real-time sensor firmware at a startup, trained GPU-accelerated models that run at the edge, built an AI sourcing agent as an internship deliverable, and shipped a live jewellery store. I like problems where hardware, data and software meet.
Outside engineering, I've run a student chapter's finances end to end, coordinated events for 1,000+ people, and helped a lean startup with hiring, onboarding and SOPs.
This site is a map because good engineering should feel a little bit like magic. It shouldn't be confusing, though, so every room is labelled.
NVIDIA Deep Learning Institute
Room II · from The Daily Commit
My three favourite builds made the front page. The rest are listed underneath.
AI agent · 2026
An end-to-end LLM pipeline built as a real internship deliverable. It turns a job description into a scorecard, finds candidates with Exa, filters them in cost-saving tiers and returns a ranked, research-backed shortlist.
Impact: Turns hours of manual sourcing into a ranked, research-backed shortlist, with tiered filtering to keep AI costs low.
Machine learning · 2025
GPU-accelerated intrusion detection on 2.8 million network records: 98.22% accuracy and 0.9989 AUC-ROC, exported to ONNX at a sixteenth of the size and 0.014 ms per sample, ready for the edge.
Impact: Real-time threat detection that's small and fast enough to run on edge devices.
E-commerce · Live
The live store for Riaura Jewellery, built hands-on in JavaScript. I learned the language by shipping it. In progress: a full PERN rebuild in TypeScript with one-time-code login, Razorpay checkout and an admin panel.
Impact: A real store serving real customers, built and shipped end to end.
Everything here lives on my GitHub.
Room III · from the Book of Spells
Every skill on my CV, written up as a spell, with where I've cast it. Turn the pages with the arrows, your keyboard, or a swipe.
Room IV
Got a project, a role, or a riddle? Pick a letter and send an owl. I answer faster than most owls fly.
Hold the mouse button and draw anywhere. Or simply type the spell's name: lumos, nox, accio, ascendio.