We automate one high-volume task end to end.
We measure candidate setups on your data (which open-source base, what post-training, what retrieval or rules around it), build the one that scores best, then run it for you or hand it over, your choice.
How we work
One task at a time
Each engagement automates a single high-volume workflow end to end.
Handover or managed, your choice
Run it with your team or have us operate it. You keep the weights, your data stays in your environment, and you can take the work in-house at any time.
A compounding pattern library
Every engagement adds task and failure patterns we carry forward, so the work gets faster and better-grounded.
Re-runnable numbers
Every number we publish comes with the data, the code, and the hashes to re-run it.
Founded by Philip Stevens
15 years in applied ML. Production work at Agoda building personalization and recommendation systems at scale, and at Quantcast managing the end-to-end ML lifecycle for core targeting models: feature engineering, model architecture, and domain drift monitoring.
- Fine-tuning (LoRA, QLoRA, full)
- Eval design & regression harnesses
- RAG pipeline hardening
- DPO alignment
- Agent workflow design
- Inference optimization
- MSc Computer Science, Univ. of Auckland
Get each new head-to-head when it publishes.
The public head-to-head is live and grows as new runs land. Leave your email and we’ll send the next one: methodology, failure analysis, per-task numbers. Technical content only.
Head-to-head releases only; unsubscribe anytime.