Four AI products, four small teams, four tight deadlines. Everything in this repo started as a hackathon submission on lablab.ai.
Hackathons are the right environment for trying ambitious ideas. You get a week, a team of strangers, and a problem statement — and at the end of it you either have a working prototype or a postmortem. Both are useful. This repo is the index of the four projects I've shipped through lablab.ai so far, with links to the live demos and writeups.
RAISE YOUR HACK · Team: Binary Ninjas
A multi-agent system that makes everyday Pakistani legal help accessible to millions of Urdu speakers. The platform combines a Retriever Agent (fetches relevant clauses), Parse & Simplify Agent (explains them in simple Urdu), Risk Advisor Agent (flags hidden risks), and a Draft Notice Agent (auto-generates legal notices) — all over local Urdu legal texts like the Rent Act and Contract Act.
Built for small landlords, tenants, and small businesses who can't afford a lawyer but still need to understand what they're signing. All conversations are saved in a reusable user profile (stored as a knowledge graph), enabling personalized follow-ups.
Stack: Llama · Groq API · LangChain · vector embeddings · Streamlit
Co-Creating with GPT-5 · Team: Code Commanders
An interactive AI tutor that teaches users to think like a programmer, not just memorize syntax. Built around a two-step learning loop:
- Learn tab — structured curriculum with bite-sized Python lessons. If you get a question wrong, an Adaptive AI Teacher re-explains the concept in a completely new way.
- Practice Arena — a library of coding challenges. Stuck? A Socratic Hint System analyzes your code and asks guiding questions instead of giving you the answer. Once you solve it correctly, the AI Code Coach reviews your working code, refactors it to a professional Pythonic standard, and explains every improvement line by line.
Designed to take a beginner from zero to writing code like a seasoned professional.
Stack: GPT-5 · Python · Streamlit · structured prompt engineering
AIstronauts: Space Agents on a Mission · Team: AI Astronauts
An AI system that monitors network performance in off-grid schools and predicts outages or hardware failures before they happen. Includes:
- Predictive maintenance — ML models that flag failing hardware ahead of time
- Transparent procurement optimization — recommends cost-effective, sustainable equipment choices
- Energy-efficient network management — load balancing tied to renewable supply
- Intelligent bandwidth management — for shared classroom connections
Built to give administrators of off-grid schools the same operational visibility that urban IT teams take for granted.
AI for Connectivity Hackathon · Team: Dreamscape Agenz
A second take on the off-grid-school connectivity problem, with a different team and a sharper focus on bandwidth optimization. Same core mission: keep the internet on, at the lowest possible cost, with the smallest possible energy footprint.
- Events attended: 4
- Submissions made: 4
- Badges: Builder · Team Player · 5× Hacker · First Steps
- Level: Hacker (525 pts)
- Profile: lablab.ai/u/@safima__
Three reasons. First — the deadline pressure strips away analysis paralysis. You ship or you don't. Second — the team composition is randomized in a way it never is at work, which makes you a faster communicator. Third — the judging panel forces you to articulate the why of your project in 90 seconds, which is exam-grade training for everything else you'll do in tech.
Also they're fun. That matters.
Saman Fatima — Kaggle Grandmaster, data scientist from Pakistan. More work on Kaggle · LinkedIn · lablab.ai.
If you're looking for a teammate for an upcoming hackathon — especially anything agentic, multilingual, or with a social-impact angle — get in touch.