- 🔭 Researching Urdu Document Intelligence — a VLM (Qwen3) + fine-tuned ByT5 pipeline for extracting and correcting Urdu text from scanned documents
- 🧩 Building Specter, an AI-powered RAG-based legal assistant for Pakistani lawyers (Final Year Project, MPVIR Group)
- 🤖 Designing Multi-LLM RAG systems with user-selectable models, embeddings, and vector databases
- 🌐 Full-stack development across MERN and modern web stacks
- 🎓 Active in IEEE RAS, GDGoC, GitHub Campus Experts, and Amnesty International at COMSATS
- 💬 Focused on NLP for low-resource languages (especially Urdu) and legal tech
- 🧠 Design and fine-tune ML/NLP models for real-world document intelligence tasks
- 🔍 Build Retrieval-Augmented Generation (RAG) pipelines with multi-model and multi-vector-store support
- 🌐 Develop full-stack web applications with scalable backend architectures & REST APIs
- 🗄️ Engineer database-driven systems (SQL & NoSQL) with clean, normalized schemas
- 🔐 Implement authentication & secure systems (JWT, structured access control)
- 🛠️ Write clean, maintainable, production-oriented code
- 🇵🇰 Urdu Document Intelligence — VLM-based (Qwen3) Urdu text extraction with ByT5 fine-tuning for OCR error correction
- ⚖️ Specter — AI-powered RAG-based legal assistant for Pakistani lawyers (FYP, MPVIR Group)
- 🔀 Multi-LLM RAG System — configurable RAG pipeline supporting multiple LLMs, embedding models, and vector DBs
- 📰 Clickbait Headline Classifier — ML classifier trained on a Kaggle dataset
- 🏪 Arthi Management System — normalized MSSQL schema with stored procedures for commission-agent business management
- 🚗 StepRides & IEEE Society Website — full-stack MERN applications
Full contribution graph, pinned repos & languages are visible natively on my profile below.