Principal Machine Learning Engineer - Clinical AI

Sanome is at a defining moment, our platform MEMORI is regulated, embedded and the 1st use case (predicting infections and downstream complications like sepsis) is live in clinical practice. The opportunity now is to scale this into a clinical intelligence platform with 100s of clinical AI models.

This is a hands-on building role. You’ll be in the code, data and models every week - architecting, training and validating clinical AI yourself, and setting the technical bar for how it is built here through design review, code review and mentoring. You’ll design, train, and validate across many different clinical data modalities and settings fusing whatever the problem demands.

Every clinical AI model has to earn its place in a clinician’s day, answering the who, what, when and why. Done well, that means catching deterioration and risk earlier, across more of the hospital and into the community, for millions of patients.

What you will do

Build and own the platform - Design, train, and validate production clinical AI across the portfolio. Own the reusable, multi-modal infrastructure and explainability layer that ships fast and safely. With Product and Clinical, choose what to build next, weighing clinical value, data, and regulatory burden. With QARA and Clinical, lead validation to prove each model meets its intended use, then work with engineering to get it production-ready.

Monitor and maintain - Own model performance once deployed. Drift, bias and fairness monitoring, local calibration, continual learning and domain adaptation. Understand how this feeds our PMS/PMCF obligations and think about PCCP.

Set the technical bar - You are the benchmark for how clinical AI is built here. Setting technical direction, making the calls on architecture and modelling approach, and raising the standard through design review, code review and mentoring.

• have demonstrable experience building clinical AI within a regulated medical device context. You obsess about continuous validation across pre-, silent and post-deployment, you know it’s not about AUROC on a test set, and you deeply understand how intended use shapes model development and validation.
• have startup experience or experience of working in an early-stage company
• are a hands-on builder who has taken models to production and can evidence it. Tell us what you shipped, who used it, and what happened to it after launch.
• are a strong engineer with strong programming and algorithmic skills in Python. You write high-standard code, build reusable platforms rather than bespoke one-offs, and own the MLOps around them - reproducible pipelines, experiment tracking, model registry and versioning, with tools such as MLflow.
• obsess about clinical utility. You want your models used at the bedside, not admired in a paper.
• have advanced knowledge of deep learning and transformer-based architectures, including scaling and fine-tuning over large-scale data, and are strong in multi-modal modelling across time-series, structured and categorical data, including free-text notes through clinical NLP (such as Clinical BERT-type models and LLM-based extraction and summarisation of clinical narrative).
• have a PhD / MSc in a related field, or equivalent industry experience

Bonus points if you

• communicate well. You can hold a room of clinicians and explain a modelling trade-off to a non-technical audience.
• have experience with survival analysis methodologies.
• have NHS deployment or health-data-partnership experience, or have worked across more than one clinical setting such as ward, ICU and community.

Competitive salary, meaningful EMI options with real upside, private healthcare, and more. The real reason you want to work here is to make a difference and be part of something that will positively impact millions of people worldwide.