Andrej Karpathy
nanoGPT, llm.c, and Zero to Hero — language models built from first principles.
2 evidence links
← Discover
Production hardening
Hire for the path from a notebook to a service
Latency, cost, observability, and inference that stays up.
9Published profiles in this specialty
100%Of those rows attach evidence
8Capabilities in the taxonomy
Why hire through this index
Production hardening is LLMOps: serving, caching, routing, and the dashboards that show when quality or spend slipped. It is listed separately from “we demoed it on a laptop.”
Available production hardening
9 published profiles listing this capability.
Chip Huyen
Author of Designing Machine Learning Systems and AI Engineering.
2 evidence links
Erik Bernhardsson
Modal founder; Annoy and Luigi — vector search and pipeline plumbing.
2 evidence links
Eugene Yan
Applied scientist; the patterns catalogue for LLM-based systems.
2 evidence links
Hamel Husain
Independent ML consultant; the field notes on LLM evaluation.
2 evidence links
Jeremy Howard
fast.ai co-founder; the fastai library and Deep Learning for Coders.
2 evidence links
Shreya Shankar
DocETL, and research on operationalising LLM pipelines over messy data.
2 evidence links
Simon Willison
Datasette and the llm CLI, plus a public archive of prompt-injection notes.
2 evidence links
Jules ExampleDemo
Agent orchestration that fails closed and leaves a trace.
Berlin · Open to work · 2 evidence links
Frequently asked
Do I need this if I only call a hosted API?
If the feature has a latency budget, a cost budget, or an on-call rotation — yes. Hosted APIs still fail, throttle, and drift.