Vipul Jain
Zi Humana Research — inventing what's next in leakage-free point-in-time feature serving, actuarial machine learning, and industry optimization platforms. Principal author of FEATSRV and FeatLock 1.0.
0009-0008-4068-4079Recommended for you
About the author
Vipul Jain is the main researcher, research programmer, and research scientist behind Zi Humana's point-in-time feature serving programme. His work connects actuarial data science with production ML systems — ensuring every online and offline feature value is temporally valid at the event timestamp it was requested against.
As principal author of FEATSRV, he established leakage-free PIT join foundations, dual Feast/Dask and PySpark engines, FastAPI serving, Redis materialisation, and the Nimbus benchmark used to evaluate underwriting, pricing, and fraud models.
Contact: zi@zi-us.com · ORCID 0009-0008-4068-4079.
Key contributions
- AuthorMain researcher — FEATSRV / Point-in-TimePrincipal author of the leakage-free actuarial feature-store research programme and Zenodo engineering report.
- SystemsResearch programmer — FeatLock 1.0Designed and shipped the interactive feature-serving console, REST surface, and same-origin API proxy workabilities.
- ScienceResearch scientist — temporal ML integrityFormalised PIT as-of join semantics, SCD Type 2 leakage dynamics, and dual Feast/PySpark equivalence checks.
- PlatformNimbus actuarial benchmarkSeven synthetic service packs with intentional temporal traps for underwriting, pricing, and fraud evaluation.
- APIsOnline Redis + offline PIT servingSub-10 ms online lookups, /pit-training, /leakage-audit, and /feature-importance for production workflows.
Top research capabilities
Zi Humana advances smarter industry systems through leakage-free feature serving, cold-chain logistics prediction, and quantum-ready optimization — with Vipul Jain as lead author on the FEATSRV programme.
FEATSRV™
Point-in-time feature store research platform for actuarial ML — Feast, Redis, PySpark, FastAPI.
FeatLock 1.0™
Live console for lineage, leakage audit, importance, training sets, and online lookups.
Leakage science
Temporal integrity checks that prevent future information from entering training rows.
Online Redis
Sub-10 ms policy and claim feature fetch for production scoring paths.
ColdMesh 1.0
Organ logistics prediction APIs — dwell, travel, and locker diversion under ischaemic constraints.
Open science
Zenodo archives, ORCID identity, GitHub research software, and developer documentation.
Smarter impact powered by Zi Humana
Leakage-free PIT joins keep actuarial training sets temporally honest at portfolio scale.
Operators inspect lineage, audits, and online Redis payloads in one research console.
Seven packs encode intentional SCD Type 2 traps for reproducible leakage evaluation.
Quantile cold-chain models feed NP-hard organ routing under hard ischaemic deadlines.
Persistent researcher identity for Vipul Jain across papers, software, and citations.
Latest work — papers, APIs & documentation
Point-in-Time Feature Serving in Actuarial Data Science: Foundations, Leakage Dynamics, and FEATSRV
Leakage-free PIT as-of joins for SCD Type 2 insurance entities — Feast, Redis, PySpark, FastAPI, and the Nimbus actuarial benchmark.
FeatLock 1.0 — Feature Serving Console
Overview, lineage, leakage audit, feature importance, PIT training, and online Redis policy/claim lookups.
FEATSRV engineering report on Zenodo
Full research report documenting dual Dask/PySpark engines, REST surface, and intentional leakage traps.
StochVRP-Mesh — NP-hard organ transplant logistics
Quantile dwell/travel prediction fused with classical VRPTW and QAOA pathfinding under ischaemic cold-chain deadlines.
ColdMesh 1.0 — Organ logistics prediction console
Dwell, travel, and locker diversion models with cold-chain risk synthesis for time-critical organ delivery.
ColdMesh API documentation — fundamentals
Authentication, endpoints, payloads, and cold-chain risk synthesis for integrators.
Stay connected
What's New at Zi Humana — Vipul Jain research releases, FeatLock updates, and open-access papers.