Junior AI Consultant at HCM GROUP – Rabat

Built an end-to-end HR chatbot using Retrieval-Augmented Generation (RAG) to improve answer accuracy by dynamically fetching HR policies and documents before generating responses. Optimized workflow orchestration with LangGraph, enabling complex multi-step reasoning, retrieval-augmented loops, and stateful conversation management. Accelerated inference speed by 60% by integrating vLLM for efficient GPU utilization, dynamic batching, and low-latency response generation. Engineered scalable APIs using FastAPI and deployed an intuitive Gradio chat interface. Generated HR Q&A pairs with LLM and validated outputs using relevancy, faithfulness, contextual accuracy, correctness, and metrics such as BERTScore. Collaborated with HR and AI teams to refine requirements, debug issues, and enhance retrieval accuracy through iterative testing and prompt engineering. Deployed the chatbot on OVHCloud using NVIDIA A100 GPU.