New playbook: when your own AI model pays. Read
Heidi Health (healthcare, ~300 employees) reported this on 22 Oct 2026 in vendor case with a named customer (source). It belongs to the learning loop layer.
Moved its clinical scribe from closed models to open models post-trained on Fireworks, and post-trained NVIDIA Nemotron 3 Ultra (SFT, then RL with a clinician-informed reward) for its 'Evidence' clinical Q&A product, now in final evaluation ahead of production rollout.
Scribe latency fell from 25 s to 7 s (3.5x), with POC to production in 4 weeks; post-trained Nemotron 3 Ultra wins ~54% of live clinician comparisons against the frontier model in production (base: ~42%).
The number covers scribe note generation (July) and the Evidence product (October), internal and live clinician evaluations.
“With a closed model the fix is a prompt edit. With weights we own, we can change the model,” Heidi Health said.
Reviewed corrections are kept as test cases and later used to train models.
Read the guide to the learning loop.
96%
lower estimated serving cost for one agent
A daily continual-learning loop for Sidekick's GraphQL agent: an expert-calibrated LLM judge scores production conversations, a self-healing pipeline repairs failures, and the repaired trajectories are trained into a smaller model (SFT, then GRPO against the judge) that replaces the frontier model (source).
13.8x
lower cost per task, with higher accuracy than the best frontier model
A Qwen3-235B model fine-tuned on Tinker with investor-labelled data to do six daily information-filtering tasks (e.g. which news matters to a macro investor) (source).
1% → 99%
valid structured output after fine-tuning a small model
A proof of value on Red Hat OpenShift AI: GPT-4.1 generated synthetic training data, then small models (Phi-4 Mini 3.8B, Qwen 2.5 1.5B/7B) were fine-tuned for multilingual structured extraction from social media, deployable locally (source).