Founding Technical Lead – Healthcare Human Data (Frontier AI), northampton
Client:
MAKZ
Location:
northampton, United Kingdom
Job Category:
Other
EU work permit required:
Yes
Job Views:
4
Posted:
08.03.2026
Expiry Date:
22.04.2026
Job Description:
We are building a healthcare-focused human data company designed to support frontier AI labs in improving model reasoning, safety, and evaluation performance.
This is not an annotation startup.
We are looking for a founding technical lead who has already worked inside the frontier AI ecosystem and understands how model training pipelines actually consume human data.
Requirements:
You must have direct experience working with a frontier AI lab (e.g. OpenAI, Anthropic, Google DeepMind, Meta AI, etc.) on:
- RLHF or RLAIF
- Red teaming
- Alignment or reasoning improvement
What You Will Own:
- Design of high-signal healthcare datasets targeting real model failure modes
- Construction of structured correction artifacts (not generic Q&A)
- Development of scoring rubrics and evaluation frameworks
- Creation of the initial validation dataset used to secure commercial contracts
- Definition of training standards for clinician contributors
- Technical credibility with frontier labs
You will be setting the bar for what “high-quality healthcare human data” means.
The Profile We’re Looking For:
- Consultant-level Doctor (or equivalent senior clinical experience)
- First-hand exposure to how frontier models ingest, filter, and score human data
- Strong intuition around model blind spots (hallucination, overconfidence, shallow reasoning, calibration errors)
- Able to think like both a clinician and a model evaluator
- High agency, startup mentality
Compensation Structure:
- Competitive consulting rate
- Performance-based upside
- Potential equity pathway based on validated impact and ARR milestones
Why This Role Is Different:
Most healthcare AI data companies compete on volume.
We compete on signal.
If you understand the difference – and have operated at that level – this will be obvious to you.
#J-18808-Ljbffr…