Research Product Manager, Fine-tuning

About the role

You will own two connected but distinct areas within the Lila product and research stack. First, you will decide which new scientific domain capabilities the Lila model should acquire and how to construct them through the right mix of reinforcement-learning environments and supervised fine-tuning (SFT) data. Second, you will own the roadmap and end-to-end delivery of fine-tuned Lila model variants for priority platform and commercial use cases, including the training data, evals, and release criteria required to ship them. These capability priorities will also feed the AI Research team's core-model training cycle as a structured input, while that cycle remains outside the direct ownership of this role. You will sit within the Research Product Manager function and work side-by-side with AI Research, joining their sprint planning and standups, to ensure the right capabilities are built and upstreamed into the Lila foundation model. You will translate ambiguous scientific and commercial pressures into clear product constraints and sequenced deliverables that the research team can execute against.

Key facts

What you'll do

• Own and maintain the prioritized backlog of training-capability packages, combining reinforcement-learning environments and SFT data, while weighing Lila's Target Product Profiles, customer and commercial needs, and quarterly model training requirements.
• Own the roadmap and delivery of fine-tuned Lila model variants for priority platform and commercial use cases, defining the required training data, identifying strategic data gaps, designing evals, and establishing release criteria.
• Coordinate the handoff of validated fine-tuned models to the app and platform teams to ensure smooth integration and downstream usability.
• Serve as a structured input into the AI Research team's core-model training cadence, which sits outside this role's direct ownership, aligning capability priorities with model-training cycles.
• Partner with AI Research, including training-pipeline owners, evals, and data-mix leads, to translate prioritized capabilities into clear, well-specified requirements and experiment designs.
• Represent customer- and commercial-driven capability requests, converting them into concrete capability asks that the research team can execute against with measurable success criteria.
• Integrate into AI Research sprint planning and standups to track delivery progress, surface risks early, and unblock cross-team dependencies between product and research.
• Define and track release and acceptance criteria for new capabilities in partnership with the evals workstream, so shipped work can be validated in controlled settings before informing fine-tuning or Lila model training.
• Maintain a clear line of sight between evolving product needs and the research and training workflows, ensuring that capability decisions are traceable to customer outcomes and model performance.
• Act as the primary product owner for fine-tuning initiatives, making decisions on scope, trade-offs, and sequencing to maximize impact across platform and commercial workflows.

Requirements

• Experience as a Product Manager on ML/AI products, ideally with exposure to LLM training, fine-tuning, or reinforcement-learning-based capability development.
• Experience owning delivery of a model or product variant for external or commercial customers, including responsibility for defining training data, evals, and release criteria.
• Demonstrated ability to build and defend a prioritized backlog against competing scientific, research, and commercial demands in a fast-moving environment.
• Strong cross-functional operator, comfortable partnering closely with a research organization by joining sprint planning and standups rather than only handing off requirements.
• Track record of translating ambiguous or evolving priorities into a clear, sequenced set of deliverables with measurable milestones.
• Comfort working with incomplete information and making timely decisions in the context of scientific exploration and product uncertainty.
• Ability to communicate effectively with both technical and non-technical stakeholders, bridging product strategy with research and engineering constraints.
• Commitment to rigorous experimentation, defining clear hypotheses, and using eval results to guide product and training decisions.

Nice to have

• Familiarity with capability-based training approaches for LLMs, including RL environments, SFT data generation, and eval design.
• Experience partnering with ML/AI research teams on eval design and release criteria, ensuring alignment between research experiments and product requirements.
• Background in a scientific or technical domain relevant to Lila's mission, which may help in specifying capability requirements and interpreting model behavior.

Practical notes

This is a full-time position based in Cambridge, Massachusetts, USA, with the option to work from San Francisco, California, USA. The engagement is full-time, and compensation includes a competitive base with bonus potential and equity, as noted in the benefits summary. U.S. benefits include medical, dental, and vision coverage; employer-paid life and disability insurance; generous paid time off; parental leave; educational assistance; commuter benefits; and a company-subsidized lunch program. International employees located outside the United States will receive a comprehensive benefits package tailored to their region, with USD salary ranges applicable only to U.S.-based roles. This role will integrate closely with AI Research sprints and planning, requiring availability to participate in regular research meetings and cross-team coordination.