Overview

Tonal is looking for a Staff Machine Learning Engineer to join our Sensors & Intelligence team and help build foundation models of human movement and physiology. You'll work where machine learning, biomechanics, physiology, and computer vision meet. You'll have access to the largest strength training dataset in the world: millions of workouts captured through cable force and position sensors, cameras, and wearable trackers.

Your job is to turn that multimodal data into general-purpose representations of how people move, exert effort, fatigue, and adapt over time. Those representations will power the next generation of Tonal's intelligence: adaptive coaching, form feedback, personalized progression, and advanced musculoskeletal health models. This role combines foundational research with work that reaches our members, and you'll help shape that balance as the team grows. You'll contribute to defining ML research directions, run rigorous experiments, and publish your work, while partnering with engineering to bring breakthroughs into the product.

This role is remote-first, based in Toronto, Ontario. We are open to candidates based in Canada and the US.

What You Will Do

Lead research and development of multimodal foundation models that learn from synchronized sensor, video, and physiological signals

Design self-supervised pretraining strategies for long, heterogeneous time series data

Develop sensor fusion approaches that handle different sampling rates, sensor noise, synchronization error, and missing modalities

Advance computer vision methods for 2D and 3D human pose estimation, motion reconstruction, and movement quality assessment

Build models that estimate biomechanical and physiological quantities such as joint kinematics, load, effort, fatigue, and recovery

Define evaluation frameworks and benchmarks that measure how well models generalize across users, body types, movements, and equipment conditions

Validate model outputs against gold-standard references such as motion capture, force plates, and clinical-grade physiological sensors

Work with ML Engineering to adapt research models for cloud deployment and on-device inference

Publish regularly at top venues and represent Tonal in the research community through papers, talks, and collaborations

Mentor scientists and engineers and help set the scientific direction of the Sensors & Intelligence team

Who You Are

A proven independent researcher with a strong track record of publications in relevant venues (e.g., NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, IMWUT/UbiComp, CHI, IEEE TBME, J. Biomechanics, or similar)

PhD in Computer Science, Electrical Engineering, Biomedical Engineering, Biomechanics, or a related field strongly preferred; we also welcome candidates whose research experience and publication record demonstrate equivalent depth

Deep expertise in at least one of the following: sensor fusion, biomechanics, physiological signal modeling, or computer vision for human motion

Hands-on experience designing, training, and debugging transformer architectures, including attention mechanisms, positional encoding for time series, and multimodal tokenization

Proven ability to train and develop ML models end to end in PyTorch, JAX, or TensorFlow, from data curation through experimentation and analysis

Strong grounding in signal processing, including filtering, calibration, time alignment, and frequency-domain analysis

Rigorous experimental design: you know how to build ablations, avoid leakage across subjects, and tell real gains from noise

Ability to independently define and drive ambiguous research problems toward measurable product impact

Clear communicator who can explain complex findings to scientists, engineers, and product leaders

Extra Credit

Experience pretraining or scaling foundation models, including distributed training and an understanding of scaling behavior

Familiarity with musculoskeletal modeling tools (e.g., OpenSim, MuJoCo, MyoSim) or physics-informed ML

Experience with wearable physiological signals such as IMU, PPG, HRV, or EMG data