Taha Hasan

I’m a Software Engineer focused on LLM-powered GenAI applications, combining NLP, Retrieval-Augmented Generation (RAG), and multimodal AI to build systems that are both accurate and dependable. I care deeply about production quality—especially evaluation, prompt iteration, and context-grounded generation.

At Mavvrik, I helped develop a production-grade Text-to-SQL agent and improved SQL generation precision by 35% through backend integration, prompt tuning, and iterative testing. I also evaluated GPT and Gemini agents to identify hallucination patterns, fine-tuned prompts for better results, and reduced query errors by 20% by analyzing 6+ Text-to-SQL frameworks.

I build end-to-end tools with modern AI and web stacks, including a full-stack RAG Assistant using FastAPI, LangChain, and ChromaDB, which reduced hallucinated responses by 90% via strict context-grounded generation. As an AI/ML Intern, I ran EDA on 30k+ data points and trained LSTM models for Bitcoin prediction (under 4% RMSE), while my education in Computer Science keeps me grounded in strong fundamentals.