AI Data-Enrichment Service

Built the scalable backend AI enrichment service that became the core of the company's API enrichment product line. Hackathon prototype that turned into the multi-quarter GenAI roadmap.

A product-shaping technical result.

Problem

The company’s data product was, by 2024, increasingly bottlenecked on what it could infer from existing records rather than what it could retrieve. Customers wanted derived attributes — seniority signals, role categorizations, custom tags — that no fixed ETL could keep up with.

Design

A scalable backend AI enrichment service sitting on top of the multi-provider LLM gateway. Customers describe the enrichment they want; the service plans the prompt, routes through the gateway, applies cost ceilings and validation, returns structured outputs, and accounts the spend per customer. Built to absorb the messiness of customer-defined schemas without the ML team needing to ship a new pipeline per request.

Started as a hackathon prototype where I led the technical work. Took 2nd place at the hackathon — by the narrowest of margins, with the cofounder backing this idea — and the company subsequently picked it up as the GenAI roadmap.

Outcome

The AI enrichment idea became what the company pursued for several quarters and is central to its AI strategy. It opened a new upsellable product class — API enrichments — on top of existing contact-data revenue.

The story I tell from this one isn’t “I won a hackathon.” It’s: a hackathon prototype, well-shaped, can become the company’s product strategy.