RAG Assistant for an Investor–Founder Platform

The challenge

The platform helps founders and investors find and engage with each other. A generic chatbot was not enough: answers had to come from the platform's own business data, and every model call added running cost.

Built as AI Developer at a remote AI product agency (2026). Client details are confidential.

What I built

• Designed a retrieval-augmented generation (RAG) pipeline that retrieves the relevant business data first, so the model answers from facts instead of guesses.
• Engineered prompts and context to keep answers focused and short.
• Shipped new product features end to end and integrated REST APIs and third-party services.
• Worked directly with the client to turn business needs into features.

The outcome

• Customer-facing answers grounded in the client's own data, which made them more accurate.
• Lower token usage and fewer unnecessary model calls through prompt and context optimisation, without losing answer quality.