AI Product Manager

Dyad is seeking an AI Product Manager to lead product work on our AI features. This role fits someone who can translate ambiguous clinical and commercial problems into well‑scoped AI product work, partner credibly with our AI Platform team on the technology itself, and ensure AI features ship with solid product discipline. AI work at Dyad proceeds in a discovery‑led and often exploratory way, at a different cadence and resolution than the rest of product engineering. The role is designed with that reality in mind. It prioritises framing, evaluation, and delivery discipline in equal measure; evaluation is a first‑class product activity, not a QA hand‑off. Reporting to the Chief Clinical Product Officer, you will operate as part of the product team and in close collaboration with our BetterLetter product manager, sharing accountability for the overall product experience. This role is offered on a hybrid basis from our London office.

Core responsibilities

• Product management: Frame AI problems in terms of user and clinical value, not just architecture or capability. Define and prioritise the AI Platform roadmap in partnership with the CCPO and other stakeholders. Ensure every AI feature has a clear intended use, acceptance criteria, and evaluation plan before it enters delivery. Work with the BetterLetter PM and the Head of Regulatory so that clinical safety is a first‑class part of the development process and medical‑device requirements are adhered to.
• Discovery, framing, and evaluation: Translate ambiguous clinical information problems into well‑posed AI tasks. Define what “good” looks like in terms the team can measure: precision and recall, hallucination behaviour, coverage, clinician trust, and practice coding preferences and standards for BetterLetter. Commission and interpret evaluation work with the AI Platform team, and feed the results into the Commercial team. Decide when an AI feature is ready to ship, expand, or pull back.
• Delivery and customer engagement: Break AI features into work that BetterLetter and AI Platform can jointly commit to, coordinating sequencing with the VP Engineering and BetterLetter PM so AI features integrate without special handling. Ensure rollout, monitoring, and fallback behaviour are defined as part of each feature, not bolted on afterwards. Run targeted discovery on AI workflows and model outputs with clinical and administrative users where AI‑specific validation requires it. Observe AI outputs in real customer contexts and feed error classes back to the AI Platform team.

Requirements

• A track record of managing and shipping AI or ML‑powered features in a commercial setting is a must, with at least 3 years of product experience. We are seeking candidates who have shipped features at a scale where the AI aspects were non‑trivial: evaluation, rollout, regression, and incident response are all important parts of delivery. You should also be comfortable translating ambiguous clinical or commercial problems into well‑posed, measurable AI tasks. Experience in a regulated or high‑assurance domain is highly desirable.
• You might come from a commercial AI product background, or you might be an ML engineer or applied researcher who has moved into product with demonstrated product judgement and the ability to work within a product process. What we are not looking for is a generalist PM who has only consumed AI products; this role must be a credible product partner to AI practitioners from day one. Healthcare experience is a plus but not required.
• In product terms, a good candidate for this role will express good product judgement through shipped, measurable outcomes and be comfortable with the discovery‑led cadence of AI work without losing that shipping discipline. On a personal level, a good fit for this role will include a comfort with collaborative work with peer product managers and a focus on understanding and communicating the risks and trade‑offs associated with AI features clearly along with mitigation.
• Related technical knowledge includes an understanding of how systems based on neurosymbolic approaches, including knowledge graphs, as well as ML and LLM‑based systems, are built, evaluated, and deployed in production; and the trade‑offs between statistical, symbolic, and generative approaches. Evaluation design, grounding, and guardrail patterns: gold sets, error taxonomies, precision and recall, regression testing, hallucination behaviour, schema‑constrained generation, retrieval, and knowledge‑graph validation are all concepts that are beneficial for a candidate for this role to be familiar with, as is experience of the cost, latency, and reliability trade‑offs for AI systems at customer scale.

Benefits

• Company pension
• 25 days of paid annual leave (pro‑rata)
• Flexible hybrid working environment
• Employee Assistance Programme