Full-stack engineer, AI-forward. I ship products end to end — API, data layer, UI, deploy — and reach for models where they make the product sharper, not where they make the README longer. Islamabad, Pakistan.
focus agentic systems · retrieval (RAG) · applied CV/ML · product engineering
stack Python · TypeScript · React · FastAPI · Node · Postgres · Docker
open to open-source work, internships, and hard problems
Patches to projects I actually use, not drive-by typo fixes.
| huggingface/smolagents | Bind function arguments with CPython semantics in the local Python executor | open |
| refined-github/refined-github | hide-inactive-deployments also hides failed deployments |
merged |
| tscircuit/schematic-symbols | Add phototransistor schematic symbol | open |
Plus ~15 shipped PRs on a private product team — Kanban boards, per-user view state, permission scoping, authenticated attachment downloads, accessibility fixes.
MeetSense — meeting intelligence pipeline.
Audio/video ingest → speech-to-text → speaker diarization, built to run as composable modules. Python
Multimodal RAG Pipeline — document
intelligence over charts and PDFs, pairing vector retrieval with vision LLMs so the model sees the
figure instead of guessing from its caption. PyTorch VLMs
DocuMind — fully local RAG
for your own notes. ChromaDB + Ollama with cross-encoder reranking, so nothing leaves the machine. LangChain Gradio
DeforestAI — satellite imagery
model that detects and tracks deforestation trends over time. Computer Vision
Browser extension that simplifies the GitHub interface and adds useful features