Heart.

The context

Corporate knowledge lives across different data sources and formats. Heart was conceived as an internal AI system to unify, query, and analyze those sources, making corporate information accessible through natural language.

My contribution

At Blue, I helped design and implement Heart, contributing to a system that integrated different data sources and supported the generation of insights for business teams.

I also collaborated on other internal automation and data integration projects during this period.

Engineering approach

The project combined Retrieval-Augmented Generation (RAG) with Text-to-SQL. Together, these capabilities allowed users to query unstructured information and structured data using natural language.

The central engineering challenge was connecting different forms of corporate knowledge through a unified query experience.

What it enabled

Heart enabled natural-language queries across structured and unstructured corporate sources, alongside the generation of insights for different areas of the company.