Senior AI/ML Researcher

About the role

You will own the research lifecycle for multimodal foundation models that synthesize wearable sensor streams, language inputs, and clinical biomarkers to model human physiology and behavior. You will design and run experiments that advance self-supervised representation learning and downstream task adaptation for health-focused applications. You will translate complex physiological signals and unstructured text into scalable model architectures that generalize across diverse member populations. You will work at the intersection of deep learning and human performance, ensuring every modeling decision is grounded in real-world health contexts. You will partner closely with engineers and clinicians to iterate rapidly while maintaining scientific rigor and reproducibility. You will define technical roadmaps and contribute to architectural decisions that shape WHOOP's long-term AI strategy. You will champion ethical, transparent, and privacy-preserving practices throughout the model development and deployment lifecycle.

Key facts

What you'll do

• Architect and train large-scale multimodal foundation models that fuse wearable sensor data, textual inputs, biomarker streams, and behavioral logs to capture complex physiological patterns.
• Lead applied research in self-supervised and representation learning methods to discover robust latent structures across heterogeneous health data.
• Engineer scalable distributed training pipelines that leverage multi-node, multi-GPU environments to handle massive physiological and contextual datasets efficiently.
• Collaborate with MLOps and data engineering teams to integrate model versioning, evaluation frameworks, and CI/CD workflows that ensure reproducibility and observability in production.
• Partner with product teams to deconstruct member needs into model capabilities, transforming foundation model outputs into personalized features that drive measurable health outcomes.
• Define and influence the technical roadmap for foundation model development, balancing innovation with practical constraints of real-world deployment on WHOOP devices.
• Implement reinforcement learning from human feedback techniques such as PPO, DPO, and GRPO to refine and align post-training foundation models with member safety and performance goals.
• Establish rigorous evaluation protocols that assess model quality, robustness, and fairness across diverse user segments while adhering to privacy regulations.
• Document model behaviors, assumptions, and limitations to support transparent communication with internal stakeholders and external partners.
• Mentor junior researchers and engineers, fostering a culture of deep learning excellence, scientific curiosity, and cross-functional collaboration.
• Ensure all modeling practices align with WHOOP's standards for ethical AI, transparency, and privacy-preserving design across the data lifecycle.
• Explore novel data-efficient learning techniques that reduce reliance on large labeled datasets while maintaining high predictive accuracy for health-related tasks.
• Coordinate with clinical and scientific teams to validate model insights against physiological ground truth and real-world member feedback.
• Optimize inference latency and resource utilization to support real-time, on-device or edge-friendly deployment without compromising accuracy.

Requirements

• Hold an Advanced degree (Master's or Ph.D.) in Computer Science, Machine Learning, Electrical Engineering, or a related quantitative field, or possess equivalent professional experience that demonstrates deep technical mastery.
• Bring 7+ years of applied machine learning and AI research experience, with a proven track record of designing and deploying large-scale modeling systems in production environments.
• Demonstrate expertise in modern deep learning architectures, including transformers, state space models, and multimodal training techniques that integrate diverse data modalities.
• Show proficiency in Python and mainstream deep learning frameworks such as PyTorch and TensorFlow, with a strong grasp of numerical computing and model optimization strategies.
• Exhibit familiarity with training models across multi-node, multi-GPU distributed compute environments, including data, model, and context parallelism strategies.
• Possess strong applied experience in representation learning and self-supervised learning, including the design of pretext tasks and loss functions for physiological and behavioral data.
• Have hands-on experience with reinforcement learning methods for post-training foundation models, including PPO, DPO, and GRPO, to align model outputs with complex objectives.
• Apply solid MLOps practices, including model versioning, systematic evaluation, CI/CD for machine learning, and monitoring of deployed systems in cloud environments.
• Communicate effectively with cross-functional teams, translating complex research findings into actionable product decisions while maintaining scientific integrity.
• Show passion for WHOOP's mission to improve human performance and extend healthspan through rigorous science and technology innovation.
• Commit to relocating to the Boston, MA office if necessary to ensure seamless integration with the local team and collaborative environment.
• Adhere to WHOOP's employment eligibility requirements as verified through E-verify, in compliance with all applicable legal standards.

Nice to have

• Preferred experience with multimodal modeling that combines time-series sensor data, text, and structured biomarkers to capture holistic views of human health.
• Prior involvement in deploying foundation models in production settings where latency, reliability, and privacy constraints are critical.
• Background in health-related AI research, including physiological signal processing, clinical data integration, or digital health applications.
• Familiarity with data-efficient fine-tuning methods such as adapters, low-rank adaptations, or mixture-of-experts to scale models without proportional compute growth.
• Experience with open-source foundation model frameworks and contributions to related communities.

Practical notes

This role is based in the WHOOP office located in Boston, MA. The successful candidate must be prepared to relocate if necessary to work out of the Boston, MA office.

WHOOP is an Equal Opportunity Employer and participates in E-verify to determine employment eligibility. It is unlawful in Massachusetts to require or administer a lie detector test as a condition of employment or continued employment.