FeatLock 1.0 is the live feature-serving console for FEATSRV — online Redis lookups, PIT training joins, leakage audits, and feature importance across underwriting, pricing, and fraud models from Zi Humana Research.
Zi Humana advances smarter actuarial ML through leakage-free point-in-time serving, dual Feast/PySpark engines, and Redis-backed online feature fetch for underwriting and fraud.
Live feature-serving console for PIT training, online lookups, leakage audit, and importance.
Sub-10 ms policy and claim feature fetch from the materialised online store.
As-of joins on SCD Type 2 entities for leakage-free training-set construction.
Scan uploaded datasets for temporal traps against event timestamps.
Authenticated JSON endpoints for health, online, offline, PIT, audit, and importance.
Seven synthetic actuarial packs with intentional leakage traps for evaluation.
Redis-backed policy and claim lookups deliver production-grade online feature latency.
Dual Feast/Dask and PySpark engines produce equivalent point-in-time training joins.
Leakage audits flag feature timestamps that violate as-of event boundaries.
Feature importance panels compare attribution across underwriting, pricing, and fraud.
Same-origin /api/featsrv forwards authenticated calls to the FEATSRV feature backend.
Same-origin proxy at /api/featsrv forwards JSON and multipart payloads to the FEATSRV feature backend. Endpoints: /health, /online/policy/{id}, /online/claim/{id}, /offline, /pit-training, /leakage-audit, and /feature-importance. The console below mirrors the original FeatLock dashboard panels — overview, lineage, risk & leakage, importance, training, and online lookups.
What's New at Zi Humana — FeatLock 1.0 API releases, point-in-time feature serving research, and actuarial ML news.