Clinical AI Engineer

Heidi; Greater London; Presencial; GBP 90 000 - 140 000; Tempo integral

Resumo da oferta

Heidi is seeking a Clinical AI Engineer with a medical degree and hands-on coding to own evaluation of AI products in clinical settings. You will design rigorous evaluations, build LLM-based judges, and ensure regulatory compliance while advancing product safety and scientific rigor.

You will work with a pod of engineers and clinicians to monitor performance, report findings, and push for changes based on evidence. This role emphasizes independence, scientific communication and durable impact.

Qualificações

• A medical degree and clinical practice with independent clinical judgment.
• Academic depth: study design, statistics, reliability measures and sample size.
• Experience evaluating medical software or clinical AI and building/validating ML pipelines.
• Ability to read and modify production code in Python or TypeScript.
• Prompt engineering and AI-assisted coding are daily tools here.

Responsabilidades

• Design evaluations for greenfield products to academic standards.
• Build, validate and maintain LLM judges and rubrics at scale.
• Report agreement with ground truth and track drift in metrics.
• Run evaluation, statistics and write up results for multiple audiences.
• Oversee post-market surveillance and regulatory feedback responses.
• Publish findings and contribute to peer-reviewed literature.

Conhecimentos

Clinical evaluation

Study design

Statistical analysis

Regulatory understanding

Independent clinical judgment

Clear writing

Independence in project work

Communication with cross-functional

Formação académica

Medical degree

Ferramentas

Python

TypeScript

Descrição da oferta de emprego

We're Heidi. We're building the future of healthcare by giving every clinician the earth's finest AI Care Partner. Our platform has absorbed the administrative chaos of 175 million patient visits and supported 67 million clinical hours. Today, we support 2.8 million patient sessions a week across 190+ countries, in 110 languages and over 200 specialties. Healthcare systems are failing us; clinicians spend more time on documentation than on patients, and the human connection that makes medicine worth practicing is eroding. Our mission is simple: double the world's capacity for care and strengthen the human connection at its heart. We found product-market fit with a freemium medical scribe that clinicians love. Now, we're expanding. Every task a clinician hands to Heidi is a patient who feels more attended to, a health system unclogged, and a clinician who gets to be a clinician again. We've grown annual recurring revenue from $1 million to $50 million in two years. To go further, we've secured US$340 million: a $100 million Series C led by Blackbird, with Phoenix Court, Point72 Private Investments and Headline, alongside a $240 million growth investment led by General Catalyst's Customer Value Fund. If you want to join us in doing work worth shipping, jump in.

The role

We're looking for a doctor who has moved into AI and can code, to join Clinical AI R&D as a Clinical AI Engineer. You'll sit inside a product pod and own the evidence for its product: define the evaluation before anyone builds, run it to a standard a journal reviewer would accept, keep watching the product once it ships, publish what the evidence shows, and change the product when the evidence says to. The pod's engineers can change the product but can't judge whether a discharge summary missed something a consultant would catch. Clinicians can make that call but can't open a pull request. You do both, and you treat the gap between a model's output and good clinical practice as the thing you're here to measure and close. The loop doesn't end at launch. You stay with the product for its life in market.

What you'll do

Design the evaluation. Define evaluations for greenfield products to an academic standard: endpoints, comparators, ground truth, sample size, analysis plan. Set the bar before the result arrives. Build the measurement. Build, optimize and validate the LLM judges and rubrics that score the product at scale. Report their agreement with clinician ground truth, not their convenience, and retire the ones that drift. Run it and write it up. Execute the evaluation, do the statistics, and write it up once in a form that serves the pod, the regulatory file and an academic submission. Run post-market surveillance. Monitor output quality in production and respond to regulatory feedback with data. Our clinical safety team triages complaints and incidents; you're their counterpart in the pod, working out what their findings mean for the product and what should change. Fix what the evidence finds. Prompt and rubric changes, evaluation harness changes, code changes in the pod's codebase, shipped through the pod's normal review and release process. Publish. Contribute to peer-reviewed literature on the translational benefit of the product. Be the clinician in the room. Catch the output that reads well and is wrong. Hold the line on clinical nuance when a shortcut would ship faster.

What you'll need

A medical degree and clinical practice with independent clinical judgment. This role runs on medical nuance, and this requirement doesn't flex. Academic depth. Study design, inferential statistics, agreement and reliability measures, sample size, handling of multiple comparisons. Thesis or publication-level work, not a methods module. Experience evaluating medical software or clinical AI, including building or validating LLM-as-judge pipelines. Enough engineering to work in a production codebase. Read it, understand what a change will do, make the change in Python or TypeScript, and get it through review. Prompt engineering and AI-assisted coding are daily tools here, and they're the floor, not the ceiling. Independence. You can run a project for weeks with light direction, and you say early when it's stuck. Clear writing. Evaluation reports, surveillance summaries and manuscripts are the primary output of this role.

Nice to have

Post-market surveillance or vigilance work under a medical device framework (TGA, MHRA, FDA or EU MDR). Appetite to learn counts for more than existing depth. Peer-reviewed publications on clinical AI or LLM evaluation. Machine learning background beyond prompting, such as fine-tuning or classical model evaluation. Specialty depth in an area Heidi is moving into. An academic or clinical-site network able to host a study.

How we show up

Build for the next decade, not next quarter