AI/ML Engineer · Computer Vision · LLM Systems (RAG & Agents) · Data Pipelines
I'm an Associate AI/ML Engineer building production AI for international clients, and a 2026 CS graduate of FAST-NUCES.
- Train YOLO segmentation models and evaluate DINOv2 classifiers for vehicle damage detection
- Build Claude-powered document automation, including an AI reviewer for civil-engineering site plans
- Build multi-agent RAG systems for internal finance knowledge
- Run Apache Airflow ETL pipelines for prediction-market and country-risk data
- Care most about honest evaluation: find where a model fails, then fix it
| Project | What it is | Stack |
|---|---|---|
| Silsila | FYP: multi-agent redistribution of near-expiry stock | Prophet · Mesa · FastAPI · React · PostgreSQL |
| BridgeRoman | Urdu voice journal with Roman-Urdu transliteration and an AI agent | Whisper · Llama 3.3 · LangGraph · FAISS |
| Qwen2-VL QLoRA | Document image to Markdown VLM fine-tune (ROUGE-L 0.66) | Qwen2-VL · QLoRA · PEFT |
| Book Nook | Cozy AI bookshelf, live demo | React · Supabase RLS · Groq |
| VibeEats | Describe a vibe and get restaurants in Pakistan | Llama 3.3 · OpenStreetMap · React |
| MAE from scratch | Masked Autoencoder with ViT-B/16 | PyTorch |
More: NLP & generative models
- UrduChatBot: Seq2Seq Transformer chatbot for Urdu
- UrduToRomanUrduGenerator: BiLSTM transliteration
- genai_assignment03: Pix2Pix, CycleGAN, WGAN-GP
- Storyteller: image captioning (ResNet50 + LSTM)
- SummarizationModel: T5/BART summarization
- SpotifyGestureControl: hands-free Spotify with MediaPipe