I build AI tools, desktop automation, and whatever sounds fun enough to break.
B.Tech CSE student at GNDU, Amritsar.
I like building things end to end. The interface, the backend, the ugly first prototype, the part that should have worked but absolutely did not, and the part where I finally figure out why.
Most of what I make sits somewhere between AI, software, systems, and automation.
Blinky is the project I keep coming back to.
It is an AI tutor + your PC on autopilot that can see your screen, understand what you are doing, and help or act on your behalf.
It started with a simple question:
What if an AI could see the same interface I am looking at instead of making me explain the whole thing?
That turned into screen understanding, computer use, voice, video editing, web research, and a companion app for controlling the desktop remotely.
The part I am most proud of is the remote workflow:
phone → Blinky → PC / IDE
I can send a task from the companion app and use it to drive desktop and IDE workflows from somewhere else.
Blinky won HACKHAZARDS '26, finishing 1st among 2,657 projects.
That matters to me because the interesting part was not just getting a demo to work. It was taking a messy idea, turning it into something usable, working through the hard parts, and actually shipping it with a team.
SIGMA
A signal intelligence and DSP project for working with IQ and WAV recordings, extracting signal parameters, identifying modulation, demodulating signals, and recovering bitstreams.
HiringMates
A technical hiring and assessment platform built around coding tests, proctoring, plagiarism and AI-code detection, and multiplayer coding.
KiranaKeeper
A small-retailer platform for inventory, orders, udhaar, and WhatsApp commerce.
GNDU Attendance System
A campus attendance system built around the practical constraints of a real college rather than an imaginary one.
There are also a lot of smaller experiments sitting around my GitHub because apparently I do not know when a side project is supposed to stay a side project.
idea → build → break → learn → ship → repeat
I learn fastest when I have something real to make.
That usually means starting before I fully understand the problem, finding out what I misunderstood, rebuilding the ugly parts, and slowly turning the prototype into something another person can actually use.
I also like hackathons for exactly the same reason: a ridiculous deadline forces an idea to become a product.
C++, Python, TypeScript, Rust
React, Next.js, React Native, Tauri
Node.js, FastAPI, Firebase, PostgreSQL
FFmpeg, GNU Radio, Docker, Linux
I am less interested in collecting technologies and more interested in figuring out what a system needs and making the pieces work together.
Building: Blinky
Learning: data structures, system design, and deeper systems work
Interested in: desktop agents, computer use, developer tools, and local-first AI
Probably breaking: Linux and Wayland
I run technical sessions, work on hackathon projects with other builders, and make coding and technology content.
A lot of it comes back to the same thing:
learn something → build something → break something → understand it better
Portfolio · LinkedIn · GitHub · YouTube · Instagram · X
I like difficult ideas. Especially the ones that sound slightly unnecessary at first.