Everyone assumes structured output enforcement works. Nobody measures whether providers actually honour it.
87.6% single-call success collapses to 26.7% over ten chained calls
Emmanuel Nwanguma · Lagos, Nigeria · Open to remote worldwide
I build ML systems and the evidence they actually work.
Feature stores, lineage tracking, evaluation harnesses, canary rollback, and the models they carry. Six open-source tools shipped, and every result published with the run that broke it.
Data Science · ML Engineering · MLOps · AI Engineering
01 / Selected work
ragwell: RAG with Verified Citation Grounding
Feb – Aug 2026Every AI tool shows sources now. Almost none verify the sources actually say what the answer claims.
126× token cost difference between retrieval and long context
ml-feature-store: Lightweight Feature Store
May – Jun 2026Training-serving skew: features computed one way for training and another for serving. The failure is silent, it is common, and it makes offline metrics untrustworthy.
Sub-10ms online serving with point-in-time correct joins
A bad model reaches production and nobody notices until users do.
Degraded model detected and rolled back in under 60 seconds, no human involved
02 / What broke
Every project here shipped a measurement. These are the bugs that would have made those measurements wrong, caught before publication.
so-agent
An earlier version of Customer.name was a required non-nullable string. Eight of the ten test tickets name nobody, so models had no legal way to say "not stated", and the benchmark recorded 80% of extractions as inventing a customer name. They were not inventing: 292 of the flagged values were the literal string the field description had asked for. With the field nullable and the description asking for null, grounding is 100% on every model except the 7B one. Give a model a way to decline and it takes it.
ragwell
A substring match on money, where N5,000,000 contains 500,000, scored every correct answer to the central question as wrong. It would have inverted the headline.
03 / Latest writing