Product Lead, Foundational Models and Post-Training

Responsibilities

• Set product strategy for Abridge’s foundation-model family, balancing capability, quality, latency, cost, safety, and controllability.
• Define hypotheses, milestones, decision gates, and success metrics to move models from research experiments through shadow mode into production.
• Shape the use of de-identified conversations, final notes, clinician edits, EHR context, and care actions as training and feedback signals with Data, Privacy, Security, and Clinical teams.
• Establish frameworks for choosing frontier models, open models, prompted workflows, or Abridge-trained models based on quality, serving cost, latency, control, and strategic value.
• Partner with Evals on product-relevant capabilities, failure modes, and model-promotion criteria.
• Convert production failures, clinician feedback, edit behavior, and product needs into training priorities and measurable user and business outcomes.
• Align ML Science, ML Engineering, Product Engineering, Clinical Science, Data, Evals, and product teams around priorities, ownership, interfaces, and dependencies.
• Communicate complex research and product strategy to executives and cross-functional teams.

Requirements

• 7+ years of product management or closely related experience, including substantial ownership of ML-powered products, model platforms, or AI infrastructure.
• Track record of turning ambiguous technical capabilities into shipped products and measurable user or business outcomes.
• Strong knowledge of the modern model-development lifecycle, including data strategy, fine-tuning, preference optimization, evaluation, inference, experimentation, and production monitoring.
• Technical judgment to reason with ML scientists and engineers about training objectives, reward design, data quality, model selection, scaling, latency, and serving cost.
• Strong product judgment regarding proprietary models versus external models or conventional systems.
• Experience creating clarity across multiple teams, defining decision rights, sequencing dependencies, resolving disagreements, and maintaining execution speed.
• High bar for evidence, safety, and trust, including understanding of escalation, abstention, and human-review paths for clinical AI.
• Excellent written and verbal communication skills.

Nice-to-haves

• Experience with LLM post-training, reinforcement learning, preference optimization, distillation, model routing, or domain-adaptive pretraining.
• Experience managing portfolios spanning frontier APIs, open-source models, and in-house models.
• Experience with human-feedback or expert-annotation systems where feedback is sparse, subjective, or expensive.
• Experience shipping AI in healthcare, life sciences, or another safety-critical and regulated domain.
• Experience with clinical documentation, medical reasoning, agentic workflows, or longitudinal context.
• Experience partnering with research teams on high-compute bets requiring staged evidence before scaling investment.