Preload logo

Robotics and physical AI · YC F26

Preload: Giving Robots a Nervous System

Preload records touch, muscle activity and first-person video together so robotics teams can learn from human demonstrations.

A robotic hand and a human hand reaching toward each other.
Preload captures how people handle objects so robots can learn from those demonstrations.Preload ↗

What a camera misses

Watch someone tighten a knot and you can follow the movement of their hands. It is harder to see when they change their grip or how much tension they apply. [1][3]

Those details matter when the job calls for care around a delicate object. [1][3]

Preload builds tools that capture those details while people demonstrate a task. The team records first-person video, contact at the hand and muscle activity together so researchers can study what happened at each moment. [1][2]

Recording the grip alongside the movement

Preload’s capture setup pairs a head-mounted camera with a tactile glove and a forearm band. The camera records the demonstrator’s view. [1][2]

The glove measures contact and pressure. The band picks up electrical activity in the muscles. [1][2]

Synchronizing those streams lets a robotics team match a visible movement to the contact and muscle signals recorded at the same time. The same approach can capture demonstrations of assembly, tool use and handling delicate objects. [1][3]

What a suturing demonstration could teach

Suturing and knot tying are early research applications. A researcher recording a demonstration could follow how someone moves an instrument and when they change tension while handling tissue. [3]

Preload is exploring whether those recordings could help researchers compare a trainee’s technique with an experienced surgeon’s and give more specific feedback. [3]

The team also wants to use the data to teach robots the movements and force control involved. That work is in early validation. [3]

From biosignals to robot training

Leo Li, Surya Pandiaraju and Runming (Tony) Wang founded Preload in 2026 and joined Y Combinator’s Fall 2026 batch. The team brings experience in computational neuroscience, biomedical engineering and neural interfaces from Caltech, Johns Hopkins and neurotechnology companies. [2]

Preload is building around the full path from recording a human task to training and evaluating a robot. It works on that problem for robotics companies, AI labs and industrial teams. [1][2]

Where Preload ranks

A ranked comparison is not yet available for Preload.

View the comparisons and methodology

The bigger picture

A demonstration can teach a robot where a person moved their hand. Preload wants the same recording to show what happened when the hand made contact. [1][3]

Written by SOTA2 from Preload’s public materials and information its founders provided.

Sources

  1. Preload · Capture hardware and robotics applications
  2. Y Combinator · Preload, team and product
  3. Preload founder input · Early surgical research and demonstration use case