Hippos is building symbiotic intelligence for the human body.
We embed edge AI inside soft exoskeletons that continuously sense, predict, and respond to the body's physical signals in real time. The system combines proprietary high-frequency sensing with active intervention layers woven into textile, engineered to resist physical failure under extreme stress, from ACL tears to long-term degradation.
We start with soldiers, elite athletes, and post-surgical patients because they generate the high-signal physical intelligence data that cannot be scraped, simulated, or collected without the hardware itself. Every deployment compounds the dataset. Every dataset trains the model.
We are a team of roboticists, biomechanists, and clinicians building the first AI that lives on the human body, learns from it, and intervenes for it.
Train on the tails. Generalize to everyone.
Most human movement is ordinary: walking, stairs, normal gait. It is abundant, already captured by every phone and watch, and it teaches a model almost nothing. The signal that predicts failure lives in the tails of the distribution, the rare extremes where the body is pushed past its limit. Hippos deploys precisely where those extremes occur.
A model trained on both extremes generalizes inward to ordinary movement, the broadest population on Earth. From there the flywheel turns: every body the model reaches generates new data, every wearer makes it sharper, and a sharper model reaches more bodies still. The data source is not a dataset. It is 8 billion human bodies.
The center of the movement distribution is already collected by every phone and watch. Hippos instruments the two tails, where the data is rare, hard-won, and predictive of failure.
Hippos Deploys with the United States Army →
May 2026First defense field deployment with the 1st Cavalry Division, the United States Army's premier armored force and one of its largest at 24,000 soldiers, known across America as the First Team. These active-duty soldiers anchor the lower tail of human movement: sustained load-bearing, dismounted operations, the cumulative physical stress of field soldiering. Durability data at the far edge of human tolerance, captured on the body in conditions no lab or consumer device can reproduce.
Hippos Deploys at Elite Sports Academy SPIRE →
February 2026The academy that trained NBA guard LaMelo Ball. First sports deployment at one of the country's leading athletic programs. These athletes anchor the upper tail of human movement: peak velocity, acute loading, the cut, the jump, the landing. The high-rate moments where the knee fails, captured continuously across a full competitive season.
Detecting ACL-Deficient Kinematics In Vivo · with Rush University Medical Center
First labeled in-vivo dataset of ACL failure biomechanics on living humans. ACL-deficient knees exhibit 22% greater rotational displacement and 58% faster rotational rate than healthy controls. Model accuracy 93 to 99% distinguishing ACL-deficient from healthy kinematics on individualized profiles; 75% precision localizing the injury window within a movement sequence from raw sensor data alone. (n=8)
Longitudinal In-the-Wild Biomechanical Capture · with SPIRE Academy
50+ hours of continuous high-intensity biomechanical data across 15 elite athletes: sprinting, cutting, jumping, contact scrimmage. First longitudinal in-the-wild dataset of its kind. Zero skin irritation, overheating, or measurable performance impairment.
Valgus Reduction in Single-Leg Landing · Hippos internal study
Live-subject quantification of active intervention across 17 athletes and three intervention designs in the dominant ACL injury scenario. Peak valgus angle change reduced from 6.1° to 1.95°. Engagement within the ~60ms injury window. Faster than neuromuscular reflex.
Rotational Restraint Under Controlled Cadaveric Load · with Cleveland Clinic
Robotic ex-vivo quantification on Cleveland Clinic's simVITRO bio-robotics platform. 10 to 25% reduction in tibial rotation under constant torque across flexion angles, with angle-dependent protection increasing at flexion. No coupling into adjacent kinematic axes.
Cross-Task Kinematic Generalization · with Edge Hill University / UK Athletics
Multi-task biomechanical capture across change-of-direction, drop landings, deceleration, and gait. 10-camera Qualisys motion capture and dual Kistler force plates at 2,000Hz. Intervention halted knee flexion in 11 of 11 activations without compensatory loading at hip, ankle, or trunk in 55% of trials.
Real-Time Valgus Detection from On-Body Sensing · with UK Athletics
On-body sensor signal validated against laboratory motion capture during single-leg jump-landing. 68% reduction in valgus angle achieved on threshold-triggered intervention in the live training environment, with no external equipment.
Sub-30ms Closed-Loop Intervention Latency · Hippos internal study
Closed-loop benchmarking on anatomically-accurate synthetic knees (3D-printed bone + ballistic gel, ~90% human fidelity). 6,400Hz on-device sensing; mean deployment 30ms; 0.15% false positive rate across 54 live cartridge deployments and 4,000 computational simulations.
The institutions deploying us. The people running them, invested in us.