The public record of an AI builder , entry 001
I build AI products, then I show my work , the architecture, the trade-offs, and the unflattering numbers. Not an expert. Becoming one, in public.
00 , The premise
Most portfolios show you a finished person. This one shows you a person being built.
Every project here includes what broke, what I traded away, and what the numbers actually were. That's the difference between a portfolio and a record.
01 Selected work , three products, told properly
Not project cards.Case studies , with the trade-offs left in.
02 , The operating loop
Build.
Small, shipped and public beats big, planned and private. Every project here exists at a URL, not in a folder called final_v2.
Write.
If I can't explain it simply, I haven't finished learning it. Writing is the compile step for understanding.
Repeat.
Momentum compounds , and so does the record of it. Each loop makes the next one faster, and leaves a trail you can audit.
03 Writing , because building isn't enough
Engineers who canexplain are rare.
04 The journey , momentum, documented
Two years,measured in milestones.
18 Jun 2025
Top 0.24% , a day inside WebVeda
Six selected from 2,500+ applicants for a day inside Ankur Warikoo's startup. The selection was flattering; watching a real team make decisions at speed was the education. Recalibrated what 'ambitious' means.
2025
2nd Runner-Up , Startup Symphony, DTU
GroomAI: an AI skincare companion for men, built to poke at a stereotype. Placed third at Delhi Technological University , and taught me more about pitching a product than building one.
Jun 2026
Three real products, shipped
Prompt Optimizer, CareAI and the Sentiment Analyzer went from repos to case studies , deployed backends, live URLs, honest metrics. The difference from the streak before: these were built to be used, not just to learn.
Jun–Jul 2026
Research Intern , IIT Delhi
Summer research internship on a Government of India project: institutional data aggregation at national scale. Built automated extraction, validation and enrichment pipelines in Python , web scraping, geocoding, cleaning , turning scattered public records into datasets people can actually make decisions with.