view
top
namanipie

cs student building things that break, learning from the pieces.

scroll

what i'm doing

learning by building things that break, fixing them, and occasionally understanding why they broke in the first place.

backend + data pipelines

building and maintaining production APIs. scraping institutional portals, solving CAPTCHAs with OpenCV, deploying FastAPI services with Docker. writing the kind of code that has to actually work for real users.

full-stack applications

working across Flutter, Next.js, FastAPI, Firebase, and PostgreSQL. most of what I build touches multiple layers — from database schemas to the screen someone's actually looking at.

ml + intelligence systems

building deterministic prediction engines. question clustering, temporal decay models, laplace smoothing. using real math instead of just throwing everything at an LLM.

PythonDartFlutterFastAPI Next.jsTypeScriptPostgreSQLFirebase DockerOpenCVHTML/CSSJavaScriptGitLinux

selected work

1def predict(course, exam_type):
2 # 389 papers, 11k+ questions
3 families = cluster_questions(
4   corpus.filter(course)
5 )
6 weights = laplace_smooth(
7   families, decay(recency)
8 )
9 # strict temporal isolation
10 return backtest(
11   weights, exam_type
12 )
01
live

MarkMint

a deterministic exam intelligence platform. ingests 389+ real university papers, clusters 11k+ questions into canonical families, and predicts upcoming exam topics using temporal decay and laplace smoothing. covers 43 courses across 8 semesters. no hallucination — every prediction is backed by historical evidence.

Next.jsFastAPIPostgreSQLPython
try it live ↗ github ↗
02
maintainer

Console

a mobile app for students who'd rather check one screen than navigate an entire institutional portal. I maintain and ship updates — unified login, student portal sync, native SRM integration, and feature patches. push notifications, social space, attendance and marks visualization.

FlutterDartFirebaseFastAPI
view on github ↗
1@app.post("/scrape")
2async def scrape(req):
3 # reuse session or re-auth
4 session = validate_session(
5   req.cookies
6 )
7 # solve captcha with opencv
8 captcha = ocr_solve(
9   session.captcha_img
10 )
11 # 9 parallel data requests
12 return fetch_all(session)
03
maintainer

Scraper API

the backend powering Console. I maintain and extend the scraping logic — session handling, retry flows, parser fixes, and new endpoint support. a FastAPI service with OCR-based CAPTCHA solving, connection pooling, and 9 concurrent data fetches. dockerized.

FastAPIPythonOpenCVDocker
view on github ↗
currently building
MarkMint live
Console maintainer
Scraper API maintainer