Lead Story
Full-stack developer working across AI, Bitcoin protocol internals, and open source. B.Tech Computer Science student in Bengaluru. I pick unfamiliar domains deliberately, which is how a raw transaction parser and a voice pipeline for ambulance handoffs ended up in the same year of work.
Page 2
By Abhiraj Singh · Staff Writer
I am a B.Tech Computer Science student specialising in AI and ML, and I spend most of my time building. I pick unfamiliar domains on purpose, because that is where I learn fastest. It is how a Bitcoin transaction parser and a voice pipeline for ambulance handoffs ended up in the same year of work.
Open source is where most of the real learning happens. I have contributed to Learning Equality, CircuitVerse, and Learning Unlimited, working on bug fixes, new features, refactoring, testing infrastructure, type systems, and language migrations. The code is only half of it. Finding the right issue, working with maintainers, and shipping something a real community depends on is a different skill from writing code alone.
Before university I was School Captain at Army Public School, Barrackpore, which taught me that leadership is mostly consistency and earned trust rather than a title.
Outside tech I care about stray dog welfare and have organised vaccination drives in my community. I am always up for a conversation about AI, open source, or pointing software at problems that actually matter.
Your code broke. Here's exactly where, and why.
AI Tooling
Practice problems that fight back. Runs your code, tells you where it broke and why, then walks the execution step by step. Tracks what you have actually retained with streaks and a heatmap, not just what you have clicked through.
Say it out loud. I'll fill the form.
Voice AI / Healthcare
Paramedics hand off patients verbally, in Hinglish, under time pressure, and then have to type it all up. This turns that spoken handoff into structured patient-care-report data, and flags every field the model was unsure about instead of silently guessing. The flags matter more than the transcription: in medical data, wrong-but-confident is worse than blank.
Ask anything. I'll pick the right tool.
Agents
A support agent that decides which tool a question needs before it answers, and holds the thread across turns so users don't repeat themselves. Routing is the hard part, because a confident answer from the wrong tool is worse than no answer at all.
Every byte accounted for.
Bitcoin Protocol
Reads raw Bitcoin blocks and transactions byte by byte and explains what is inside them. No library does the parsing: SegWit witness structure, address derivation, and fee analysis are implemented straight against the protocol spec.
Ten thousand futures. One hour ahead.
Quant / Simulation
Forecasts Bitcoin’s next hour as a range, not a number. Geometric Brownian motion driven by FIGARCH volatility and Student-t shocks, ten thousand runs per prediction. Across 720 backtested hourly predictions the 95% interval held 95.1% of the time: the stated confidence is the real confidence.
Pick the coins. Do the fee math. Sign it.
Bitcoin Protocol
The other half of a wallet: pick which coins to spend, work out the fee, build the PSBT. Four coin-selection strategies and two-pass fee math, because the fee depends on transaction size and the size depends on which coins you picked, so the numbers have to settle before you can sign.
Nobody picks the winner. The chain does.
Web3 / Solidity
A lottery with no operator to trust. Entry and winner selection both happen on-chain on the Flare Coston2 testnet, so the draw is something anyone can verify rather than something someone promises.
Thousands of titles. One chart.
Data / Dashboards
Game statistics, trends, and search across the RAWG catalogue, rendered as interactive charts instead of raw rows. A database that big is useless until you can see the shape of it.
One more run. You'll beat it this time.
Arcade
A retro arcade survival game: dodge, survive, watch it speed up. Hand-rolled collision detection and a difficulty curve that keeps ramping, with high scores that persist between sessions.
Page 4 · The Sunday Edition
Page 5
What I reach for, listed the way a paper lists the markets.
Page 6
2025 to 2029
Medhavi Skills University
Coursework in Python, full-stack web development, data structures and algorithms, the mathematics of machine learning, and database systems.
Apr 2024 to Mar 2025
Army Public School, Barrackpore
91% overall, with 97 in Computer Science and 95 in English Literature. School Captain, leading the student council and running workshops, campaigns, and cultural events.
Page 8
Situations Wanted
Abhiraj Singh, B.Tech Computer Science (AI/ML), Bengaluru. Builds full-stack applications, AI tooling, and protocol-level software. Open to conversations about internships, collaboration, and open-source work.
itsabhiraj27@gmail.com