Rehan Khalil
Hey, I'm Rehan. I'm based in Karachi, and I got into AI because I liked the idea of teaching a machine to actually understand something text, images, whatever. Most of my time goes into computer vision and NLP projects, but I also freelance in web and app development on the side while I work toward doing AI full-time. I'd rather show you what I've built than talk myself up, so scroll down.
Browser-based AI platform: text-to-3D generation, an interior design module, and a virtual try-on system built on a custom CV pipeline. Solo build, solo deploy.
read case study
Problem: Wanted a single platform combining generative 3D, room design, and virtual try-on — tools that normally live in separate apps.
Approach: Built three independent modules on a shared Three.js viewer, connected to separate model backends (Tripo3D for generation, custom MediaPipe masking for try-on).
Result: Fully deployed platform, solo-built end to end, presented as final year project.
TrOCR for recognition + NLLB for translation, wrapped in a simple upload-and-translate interface. Turns scanned Urdu documents into readable English, trained and run locally.
Fully local multi-user RAG system with a chat-style query interface and per-document access control. Flask, ChromaDB, Llama 3.2 via Ollama.
read case study
Problem: Teams wanted a RAG assistant without sending private documents to a third-party API.
Approach: Built a fully local pipeline (ChromaDB + Ollama) with per-document access rules so restricted files never enter another user's context.
Result: A working multi-user system where access control is enforced at retrieval time, not just at the UI layer.
| ml/cv | PyTorch · OpenCV · Hugging Face · MediaPipe · TrOCR · PaddleOCR | 90% |
| backend | Python · Flask · Firebase · Node.js | 80% |
| frontend | React · Three.js · JavaScript | 75% |
| infra | Railway · Cloudflare · Docker | 70% |