ML Engineer, Full Stack Developer and more.
TLDR: I am an ML engineer and full-stack developer. I research how small AI models reason, and I build products end to end. I look for real gaps that people live with every day, and build whatever fills them.
I like owning the whole thing: the database, the API, the last button, and the tests that keep them honest. HostelMate, which I built alone, has over 300 automated tests (260+ Jest, 59 Playwright) and an 80% coverage floor. It's open source, with a live demo.
I keep learning on purpose and pick up whatever a problem needs. In my second year I co-authored a paper accepted at ICTCS, built Gearbox, a tool that lets expensive AI agents hand routine work to cheaper models safely, and finished in the top 5% of 35,000 teams in the Amazon ML Challenge.
I grew up in Kota, Rajasthan. In 2015 I won 1st place at the UCMAS All-Rajasthan Abacus Championship, and I've been curious about how numbers and things work ever since. My father trades grain at Bhamashah Mandi, and watching him check prices every morning is where my Mandi bot comes from.
I study B.E. Computer Science (Data Science) at BMS College of Engineering, Bengaluru. Since March 2026 I've been building HostelMate, a hostel management platform for Indian colleges, on my own. This summer I interned at Trinkerr as a software developer. Before that I ran social media for Veynqor, a US-based B2B startup, and built its website. I'm looking for SWE or ML internships.
Outside work, I'm a lifelong learner who likes finding a real gap and building something to fill it. Away from the screen, I play table tennis and snooker.
I'm looking for SWE or ML internships. I build products end to end, from the database to the last button, and I research how small AI models reason.
Tech stack: Django, DRF, PostgreSQL, Celery, Redis, React, TypeScript.
Veynqor is a US-based B2B startup. I worked with it remotely, alongside my studies.
What I build with, from the database to the last pixel.
BMS College of Engineering, Bengaluru. CGPA 8.42.
Kota, Rajasthan. Scored 80%.
Kota, Rajasthan. Scored 92%.
Indian college hostels still run on paper registers, WhatsApp complaint groups and handwritten leave forms.
One platform for students, wardens and parents: attendance, leave, complaints, mess, fees, visitors and lost-and-found.
Highlights:
My Role:
Links:
Retail investors in India see a P/E of 24 and can't tell if that's healthy, because what's normal differs by sector.
Stock research that scores every NSE company on six fundamentals against its own sector, and explains each score in plain language.
Highlights:
My Role:
Links:
Restaurants throw away good food while NGOs nearby run short.
A marketplace where donors list surplus food and NGOs or volunteers claim it before it expires.
Highlights:
My Role:
Links:
Action items get lost after meetings.
Turns meeting audio into action items, decisions and open questions, each linked to the exact quote and speaker.
Highlights:
My Role:
Links:
Hiring managers drown in candidate PDFs.
Reads résumés, normalises skills and recommends matches. The final score is plain code, so the AI can't be talked into a result.
Highlights:
My Role:
Links:
A playful one: leaving a video call is dull. What if you could vanish the way the Endgame snap does it?
A macOS virtual camera: snap your fingers and you turn to dust while the room behind you stays live; snap again to come back. Works in WhatsApp, FaceTime, Google Meet and Zoom through OBS.
Highlights:
My Role:
Links:
My father trades grain at Bhamashah Mandi in Kota and checks prices on a website every morning.
A bot that fetches today's prices for what he holds and predicts profit from incoming crop arrivals and the weather.
My Role:
Our college runs events all the time, and the tech ones get lost in WhatsApp groups.
A bot that reads the groups and tells me only about hackathons and tech events.
My Role:
A common test truncates a model's reasoning and forces an answer. If an early forced answer is right and the final one wrong, people call it 'overthinking'. But forcing the answer is itself an intervention.
A two-tier protocol that separates what the probe forces from what the model actually commits to, run on DeepSeek-R1-Distill-Qwen-1.5B, Qwen2.5-3B and Llama-3.2-3B over GSM8K and MATH-500.
Highlights:
My Role:
Expensive AI agents spend most of their tokens on small, routine subtasks a cheaper model could do, if anyone could trust the cheaper model's 'done'.
An automatic transmission for AI agents: the agent hands subtasks to cheaper models, keeps working, and accepts an answer only after it passes a real test. Failures move up to a stronger model with exactly what failed.
Highlights:
My Role:
Links:
People who lose their voice to throat or oral cancer surgery, a stroke or ALS can still move their lips, but no sound comes out. India has one of the highest rates of oral and throat cancer.
Reads a person's silent lip movements through a phone camera and speaks for them in Hindi, in their own voice rebuilt from old voice notes and videos: lips to text (an LLM fixes look-alike sounds) to cloned speech, or lips straight to speech units.
Highlights:
My Role:
Links:
AI tutors can help so much that people stop learning.
Psychology × AI research on how people learn with AI tutors.
Highlights:
My Role:
Match the same business across three messy sources with about 12.5 million records: typos, swapped legal forms, missing addresses, and names written in Devanagari, Tamil, Telugu or Gurmukhi.
A three-stage pipeline: cheap candidate search on rare name and address keys, a LightGBM matcher with about 70 features, and a decision layer that enforces one-to-one matches. A transliteration dictionary learned only from the training pairs, so Indic-script names meet their English versions. Score 0.965 out of 1, rank about 1,800 out of 35,000 teams. Built with team DayTrippers.
Won 1st place in the state-wide UCMAS abacus championship in 2015.
Completed Harvard's CS50x, weeks 0 to 9 in C, with FoodBridge as the final project.
Links:
Rated 1443 on Codeforces, plus 200+ DSA problems solved in C++.
Links: