HR Policy Assistant Prototype

Demo placeholder. Employees could not find answers inside long policy documents, and repetitive questions consumed HR service capacity.

Demo placeholder. Built a retrieval-augmented generation prototype over a small approved policy corpus: chunking and embedding documents, retrieving by semantic similarity, and constraining the model to answer only from retrieved passages with citations.

Demo placeholder. A working RAG prototype with a curated evaluation set of real policy questions, measuring answer accuracy and citation correctness before any user-facing claims.

• Python
• LLM API
• Embeddings

Demo placeholder. Replace with real evaluation results once available — this record demonstrates how a project links to GenAI concepts and an HC-AI use case.

Demo placeholder. Retrieval quality, not model choice, dominated answer quality — and an evaluation set was worth more than any prompt tweak.