Why ML Models Fail in Production: Beyond POC
A 91%-accurate predictive maintenance model built in four months sits unused six months later. The problem wasn't the algorithm—it was missing pipelines, documentation, and ownership. Enterprise ML stalls not because science fails, but because infrastructure and operations do. Full article: https://davidohnstad.net/ml-models-fail-production-poc/
Authors
- David Ohnstad
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-02
- DOI
- https://doi.org/10.5281/zenodo.23892954
- Primary Topic
- Software Engineering Techniques and Practices
- Type
- article
- Field-Weighted Citation Impact
- 0.00