About
Built from the work of building healthcare AI.
We didn't begin with a marketplace. We began by trying to make AI useful to clinicians in a specialised field, and that pushed us into data generation, clinical constraints, and finding out where convincing outputs fail.
How we got here
From a practitioner product to data infrastructure.
Building software for a specialised clinical field exposed a broader problem: specialised healthcare data is hard to produce and harder to prove. Meethealth Labs is our answer to that problem.
- Product
An AI-native platform for nutrition professionals
Clinical nutrition workflows, patient management, structured meal planning, and practitioner tools. Built for dietitians, not for patients to self-serve.
meethealth.us(opens in a new tab) - Infrastructure
Generation pipelines and structured outputs
Making the product useful required specification-driven synthetic generation, schema-bound outputs, and model comparison across several open and commercial models.
- Verification
The Assessment Gate
A deterministic, versioned contract of clinical and structural rules, run over millions of generated records. It is how we found most of the failures our research describes.
- Now
Meethealth Labs
Training data, preference data, evaluations, and custom generation for teams building healthcare AI, delivered with the checks and reports that make them auditable.
The team
Engineering and clinical practice, in the same room.
Meethealth combines machine learning engineering with clinical nutrition practice. The same team that built the generation pipelines, the deterministic Assessment Gate, and the evaluation rubrics works directly with clinical nutrition expertise when scoping and checking data.
Machine learning engineering
Generation pipelines at million-record scale, structured outputs, model comparison, and the deterministic verifier that gates every corpus.
Clinical nutrition practice
Guideline-anchored rules, rubric design, and the judgement about which failures matter clinically. Nutrition is the specialty we cover in-house.
Evaluation and research
Audits of how diet-generating models fail, written up with methods, counts, and limitations, and turned into enforceable gate rules.
Mission
Make specialized healthcare AI data more useful, measurable, and accessible to the teams improving medical AI.
Measure before you claim
Every figure on this site traces to a stored artifact: a gate run, a contract fingerprint, or a file you can download.
Label evidence honestly
Automated checks, rubric scores, and clinician review are different things, and we name each one for what it is.
Specialise before expanding
New specialties start with matched clinical experts and a feasibility review, never with a generation run.
Talk to the people building it.
We'd rather have a direct technical conversation than send a brochure.