Builder · Assam, India · Still in school
SAJURJYA
BORA

I build software that quietly earns its place. Some of it helps my family sell jewellery. Some of it trades my own money. Some of it gets me through tomorrow's physics exam.

Scroll

I spend more time deleting ideas
than shipping them.

Selected work
Commerce — Live

Kakoli's Collection

Built because my family deserved software that looks as good as the jewellery they make. A real store — catalogue, payments, an admin panel behind the counter, and a cinematic opening I definitely over-polished. Customers use it. Orders happen.

Next.js 14SupabaseRazorpay
Fintech — Live

Skew Hunter

Built because I got tired of opinions pretending to be statistics. An options desk that runs itself — three brokers, 1-second market data, risk that never sleeps — trading my own capital while I'm in class. It has scars. I kept them.

Verifiable, not vibesThree broker APIs integrated and live. Every fill is broker-confirmed before it's recorded — no phantom trades.
Runs itselfUnattended every market day, 09:15 to 15:30, with watchdogs, kill-switches and a force-paper failsafe when infrastructure degrades.
Data it earnsCaptures its own 1-second option-chain feed all day — 40+ GB and growing. Tomorrow's ideas get judged on it.
No performance theaterYou won't find a win-rate here. The sample is young and the market grades slowly. The engineering is checkable today; the edge is being earned in public.
Skew Hunter app dashboard
trade replay view
engine deck
research charts
PythonReact NativeSupabaseAWS
Ed-tech

SynaptiQ

Built because every exam app felt like homework wearing brighter colours. A periodic table you can play with, streaks, momentum — made between the same chapters it teaches, by someone actually sitting these exams.

Next.jsFramer MotionCapacitor
Research

The machine that tells me no

Built because I kept believing my own backtests. Every idea faces real premiums, real costs and data it has never seen — before a single rupee moves. Most ideas die here. That's the job.

Most expensive lessonBeautiful backtests lie. My best ever made ₹10.8 lakh on paper — one hidden lookahead bug. Real value: zero.
Humbling bugI once "fixed" a leak and the engine stopped trading entirely. The thing I deleted was the mechanism.
Current dataset40+ GB of 1-second option ticks, self-captured, growing every market day.
Rejected vs shippedHundreds of ideas vs a handful. The ratio is the product.
Pythonpandaswalk-forward validation
Dev tool — Open source

DevDoctor

Built because every README eventually lies. A zero-config CLI that catches documentation drift: broken links, stale version numbers, flags that no longer exist, dead imports in code examples — nothing executed, no network needed. First real catch: broken links in dotenv's README, a package downloaded 120 million times a week.

Node.jsCLIzero-config
Current obsessions
Making an options strategy survive data it has never seen — not just the week it was born in.
Working out how much of building software is actually deciding, and how little is typing.
Educational apps that don't feel like homework.
Shipping real products while finishing school, without either one noticing.
About

I architect, direct, debug and validate. AI does the typing.

Every project starts as a picture in my head and a page in a notebook. I break it into decisions, direct the build through every one of them, hunt the bugs, and test the claims — and I reject anything that doesn't match the picture. The typing got fast. Knowing what deserves to exist still takes me months.

First websitesa store for my familya trading enginemy own market dataa machine that rejects my ideasall of it while finishing school

Building something? Talk to me.

Currently open to freelance builds, collaborations, and conversations worth having. Replies within a day.

BuildingSkew Hunter's next validation run
Testingthe one exit rule that survived
Studyingfor boards — allegedly right now
Wonderinghow far one person and fast tools can go