Case Study: Epona Science achieves faster, scalable, reproducible ML-driven racehorse selection with Pachyderm

Epona Science - Customer Case Study

Epona Science buys, breeds and identifies elite racehorses and needed to turn messy, globally sourced data — from genomes and x‑rays to track records and gait measurements — into reliable inputs for sensitive machine‑learning models. The volume and variety of imperfect genetic and image data, plus brittle manual “glue” scripts, meant model training and data processing took weeks or months and made reproducibility and forensic analysis difficult, so Epona Science turned to Pachyderm and its containerized, versioned data‑pipeline platform.

Pachyderm implemented containerized, provenance‑tracked pipelines on Kubernetes with autoscaling, enabling continuous, reproducible model builds and easy rollback/forensic analysis. The result: data and models that once took days or weeks to process now run in minutes, and Epona processed 10,000 new photos in a month (vs. a year previously); teams avoid costly mega‑instances by auto‑scaling pods, speed high‑stakes purchasing decisions, and maintain full data lineage — all delivered via the Pachyderm platform.

Epona Science

Ryan Smith

Head of Data Science