- π€ Senior AI Engineer building production-grade LLM systems, not just demos
- π§ 4+ years experience delivering end-to-end AI platforms (RAG, Agents, OCR, Voice AI)
- π Built 15+ real-world AI systems handling messy, unstructured, enterprise data
- π Specialized in RAG architectures, hallucination control, and hybrid retrieval systems
- βοΈ Strong focus on scalable FastAPI backends, async pipelines, and system reliability
- π Based in Islamabad, Pakistan, delivering high-impact AI ownership globally π.
- π¬ Ask me about RAG failure modes, hallucination mitigation strategies, or designing scalable LLM APIs in production.
π‘ I donβt just integrate APIs - I design AI systems that actually work in production.
---
- π· Designing Agentic RAG Systems with self-correction and multi-step reasoning
- π· Building Hybrid Search Pipelines (BM25 + Vector + RRF)
- π· Developing Document Intelligence Systems (PDF β Structured Data)
- π· Handling noisy OCR data and complex layouts
- π· Reducing LLM hallucinations with retrieval + validation layers
- π· Creating real-time AI APIs with streaming responses
- π· Deploying LLM systems on GPU infra (RunPod, local models)
| Project | Description | Tech Stack |
|---|---|---|
| Agentic RAG System | Production-ready Agentic RAG system with self-correction loops, hybrid retrieval (RRF), and multi-tenant architecture. | FastAPI, Gemini, ChromaDB, LangGraph |
| RAG Tutorials | 30+ hands-on implementations covering naive β advanced RAG, hybrid search, reranking, and agent workflows. | Python, LangChain, FAISS |
| Multi-Engine OCR | High-accuracy OCR pipeline combining multiple engines with preprocessing for real-world noisy documents. | Python, Docling, OpenCV |
| Sarah Voice Agent | Real-time conversational AI voice agent with speech-to-speech pipeline and contextual memory. | Python, FastAPI |
| Rahnuma | Urdu conversational AI assistant with custom prompts and localized intelligence. | Python, LLM |