Puneet Rawat

I am a Generative AI engineer with hands-on experience building RAG systems, multi-agent LLM orchestration, and model fine-tuning using frameworks such as Unsloth and Transformers. I focus on scalable pipeline design, LLM optimization, and prompt engineering while maintaining a builder’s mindset and fast iteration cycles.

I've developed production-style projects including an autonomous AI debate system with LangGraph and LangChain, a DeepResearch multi-hop RAG agent leveraging hybrid retrieval and web search, and fine-tuned LLaMA-3.1-8B with PEFT (LoRA) for structured information extraction from noisy product text. I integrate tools like Streamlit and FastAPI to deliver usable interfaces and services.

I bring practical experience across the GenAI stack—LangChain, LangGraph, Hugging Face, Ollama, and OpenAI—paired with data and engineering skills (Pandas, PyTorch, FAISS/Chroma, SQL). I aim to build reliable, evidence-grounded AI systems that balance autonomy, interpretability, and human-in-the-loop control.