Delphina: Building the enterprise context factory

Coding agents have gotten really good. But they still make too many mistakes when working with data. The problem isn't that they can't write SQL or perform sophisticated analysis. It's that they don't understand the business.

There's an enormous amount of institutional context required to do the right analysis. You can't just look at tables and infer everything. Maybe revenue excludes free trials. Maybe there's an old wiki explaining how a metric is calculated. The more you drill into a question, the more context you need, and mistakes compound. Worse, the answers often look completely reasonable.

AI is making it possible for business leaders to answer their own questions without waiting for an analyst or a dashboard. But 80% accuracy is a failure when you're making important business decisions.

Jeremy and Duncan built Delphina to solve this. Their enterprise Context Factory uses agents to discover, organize, and maintain the institutional knowledge needed to work with a company's data. It reads documentation, dashboards, and historical SQL queries, reconciles conflicting definitions, and turns them into context that AI agents can actually use.

Crucially, this isn't a one-time effort. Data changes, metrics change, and the questions people ask change. Delphina's agents continually identify gaps, investigate inconsistencies, and work with domain experts to validate fixes. The knowledge base improves as people use it.

Today, companies including Substack, LATAM Airlines, Coursera, and Upstart use Delphina to understand their businesses and make decisions faster, without waiting on their data teams.

But the opportunity is much bigger than analytics.

Coding agents can already build pipelines, write transformations, and debug code. What they lack is context about which data matters, what needs fixing, and what changes would break something important. Delphina has a unique view into how a business actually uses its data. That puts it in a position to take over more data engineering work: maintaining models, fixing data quality issues, and operating infrastructure based on what actually matters to the business.

The Context Factory starts by making analytics trustworthy. Over time, that same context could enable agents to build and maintain much of the data stack.

Jeremy and Duncan have spent their careers building data and AI systems at Uber, GoPuff, and Tecton. They recognized early that coding agents would change how companies work with data, and that context would be the bottleneck.

We're incredibly excited to be backing Jeremy, Duncan, and the Delphina team as they build toward this future.