In this episode of The OpenYield Markets Podcast, host Mark Hebert sits down with Miguel Jaques and Daniel Dyulgerski, co-founders of Terrapin, a Scotland-based startup rethinking how reference data is built and delivered in fixed income markets.
With backgrounds spanning multi-asset portfolio management and machine learning research, the Terrapin founders explain how they’re using fine-tuned LLMs, rigorous OCR pipelines, and automated QA to achieve ~99% automation in extracting and maintaining municipal bond reference data. Together, they explore:
- What “reference data” actually means and why it underpins analytics, settlement, and trading
- Why Muni bonds, with their scale and complexity, became the perfect use case for AI
- How incumbents still rely on armies of manual typists—and why automation changes the cost base
- Building infrastructure that keeps LLMs accurate at scale
- Why flexible APIs, fast feature rollouts, and lower costs win clients from Goliaths
🔔 Subscribe for more episodes to gain insights from leaders shaping the future of fixed income trading.