Research Engineer, Post-Training

SF · NYC

About RoVR

RoVR builds the autonomous system of execution for healthcare: AI that runs clinics end to end, acts across every system they depend on, and improves with every outcome. Healthcare is one of the most complex real-world environments for autonomous intelligence, demanding continuous coordination, reasoning, and decision-making under uncertainty. We’re pushing the frontiers of agentic systems, long-horizon reasoning, memory, and reinforcement learning to build AI that pursues goals, recovers from failures, and learns from experience. We’re building the foundation for a future where entire healthcare systems operate, adapt, and improve autonomously.

Why RoVR

At RoVR, we’re on a mission to turn practices into AI-Native, Self-Optimizing Learning Health Clinics. At RoVR, we are building a multimodal, agentic operating system to launch, manage, and scale A-LHCs nationwide. We collaborate with large health organizations as well as local physician groups, modernizing legacy operations to launch these next-generation clinics.

Central to this vision are at RoVR, autonomous agents that orchestrate and optimize clinical, administrative, and patient engagement processes. By alleviating repetitive tasks, RoVRs free providers and admins to focus on proactive decision-making guided by actionable insights, laying the groundwork for a more responsive, outcome-driven healthcare ecosystem.

By uniting clinical and operational intelligence in one cohesive platform, we're building the infrastructure for a Learning Health System that can collaborate with and empower clinicians to redefine the patient experience and transform health outcomes.

Based in the heart of NYC and backed by leading venture capital firms/visionary healthcare leaders, RoVR is on a mission to transform the future of healthcare. We are seeking mission-driven engineers, designers, and operators to join our team and help us build the foundational operating system that will drive medicine for the next century.

The Role

We are looking for a product builder to own how patients experience RoVR. From finding a provider and booking a first visit to understanding next steps and staying connected between appointments, you would make complex care feel understandable and manageable.

You would also shape how patients interact with agents: when an agent should take initiative, what it should explain, when it should ask for confirmation, and how a person takes over. These decisions are central to the product.

The work combines product judgment, interaction design, and experimentation. You will spend time with patients and care teams, turn observations into prototypes, and work with engineers to ship and evaluate them. You will be accountable for whether people can understand the experience, complete what they came to do, and get the help they need.

Responsibilities

Own patient journeys across onboarding, provider discovery, scheduling, intake, communication, and ongoing care.
Study how patients and care teams use the product. Turn confusion, abandoned tasks, and repeated requests for help into concrete improvements.
Design interactions across the patient portal, chat, and voice. Define how agents communicate progress, handle uncertainty, request consent, and hand work to people.
Turn ideas into flows, prototypes, and clear product behavior. Work closely with engineering, design, and clinical operations to carry them through launch.
Run experiments with explicit hypotheses and success criteria. Measure task completion, patient understanding, effort, and follow-through.
Build experience evaluations with realistic patient scenarios, including interruptions, misunderstood requests, changing preferences, and failed actions. Use them to assess changes before and after release.
Make accessibility, health literacy, language needs, and patient control part of the product from the beginning.

Minimum qualifications

You have shipped a consumer product or service and can explain the decisions you made, what happened after launch, and what you changed.
Strong product judgment and interaction design instincts. You can make a complex process easier without hiding information people need.
Ability to turn an idea into a testable prototype using design tools, AI tools, or lightweight code. A software engineering background is not required.
Experience speaking directly with users and combining those observations with behavioral data to decide what to build.
Clear writing. You can define product behavior precisely and write language that helps someone understand what happens next.
A practical understanding of AI capabilities and limitations, including how unpredictable behavior affects the experience.
Ownership from problem discovery through launch and iteration.

Preferred qualifications

Experience building consumer healthcare products or services that involve sensitive information and consequential decisions.
Experience designing conversational, voice, or agent-based experiences.
Experience improving onboarding, activation, or ongoing engagement through measurable product changes.
Familiarity with accessibility, multilingual experiences, or designing for people with different levels of health and digital literacy.

Logistics

Location: San Francisco or New York, on-site.
Employment type: full-time.
Compensation: competitive salary and meaningful equity.
Visa sponsorship is available.

RoVR is an equal opportunity employer and does not discriminate on the basis of race, gender, sexual orientation, gender identity/expression, national origin, disability, age, genetic information, veteran status, marital status, pregnancy or related condition, or any other basis protected by law.

The Role

We are looking for a research engineer to advance how open-weight models learn to reason, use tools, and carry out complex work in healthcare. You would own research and engineering across post-training: identifying important research questions, developing methods, running experiments, and translating findings into agents that perform better in production.

You will have substantial autonomy to shape the research agenda and carry projects from an initial hypothesis through implementation, evaluation, and release. We expect you to engage deeply with emerging research, challenge existing approaches, and develop original contributions that can advance the field. The work combines scientific judgment with hands-on ownership of data, training, and evaluation.

Responsibilities

Define and drive research directions in post-training for long-horizon reasoning, tool use, memory, and recovery from failures.
Develop and evaluate training methods across supervised fine-tuning, preference optimization, and reinforcement learning. Investigate new approaches where existing methods fall short.
Own the post-training pipeline end to end: data selection and mixing, training stages, evaluation, and release.
Turn production trajectories and human feedback into training signal, including preferences, corrections, reward models, and verifiers that capture successful actions and outcomes.
Design rigorous experiments with strong baselines and clean ablations. Establish which methods, data mixtures, and base models deliver meaningful gains per unit of compute.
Stay at the forefront of relevant research. Reproduce promising results, examine their limitations, and extend them to the demands of agents operating in real environments.
Work with Evals and Environments, Data Foundry, and Infrastructure to establish reliable evaluations, improve training data, and keep runs reproducible and regressions traceable.
Produce original, publishable research and communicate findings through papers, technical reports, and open-source contributions where appropriate.

Minimum qualifications

Hands-on experience post-training open-weight models with SFT, preference optimization, or RL methods, with measurable improvements you personally drove.
Strong foundations in machine learning, optimization, and the design and evaluation of learning systems.
Strong Python and PyTorch skills, with the ability to implement research ideas, debug training behavior, and work effectively in a large training codebase.
Demonstrated ability to conduct research independently: identify a meaningful problem, formulate hypotheses, design experiments, and carry the work through to defensible conclusions.
Experimental discipline: appropriate baselines, clean ablations, reproducible results, and honest reporting of limitations and negative findings.
Ability to read research papers critically, reproduce their methods, and determine whether the findings hold in a new setting.
Willingness to own data as well as method, and to connect research results to the behavior of deployed systems.

Preferred qualifications

PhD in computer science, machine learning, artificial intelligence, or a closely related field.
Strong publication record at leading research venues such as NeurIPS, ICML, ICLR, ACL, or EMNLP, with substantial intellectual and experimental contributions.
Original research in reinforcement learning, post-training, reasoning, reward modeling, or learning for tool-using agents.
Experience with distributed training, large-scale experimentation, or improving training efficiency.
Experience training and evaluating agents on long-horizon tasks with sparse or delayed feedback.
Experience working with sensitive or regulated data.

Logistics

Location: San Francisco or New York, on-site.
Employment type: full-time.
Compensation: competitive salary and meaningful equity.
Visa sponsorship is available.

Apply for this Job