# Parametric

> Parametric is a San Francisco robotics company building autonomous physical businesses - operations that run themselves. Starting with a wash-and-fold laundromat, its bimanual mobile robots use reinforcement learning to fold, sort, and handle laundry, learning new behaviors on-site from customer feedback in under an hour. The team applies frontier-lab techniques - dense reward models, an RLHF-style feedback loop, and interpretability tools borrowed from language research - to physical work, reporting 3x higher task reliability than leading baseline models on matched hardware.

- **Founded:** 2025
- **Headquarters:** San Francisco, United States
- **Founders:** Cody Swain (Co-Founder & CEO), John Newsom (Co-Founder & CTO)
- **Team size:** ~7-12
- **Products:** Autonomous wash-and-fold laundromat, Bimanual mobile robot, RL feedback pipeline (RLHF for robots)
- **Notable:** Reported 3x higher task reliability than a state-of-the-art baseline model on matched hardware and benchmarks., Reported robots beating Physical Intelligence's Pi0.5 model with 20% greater speed on matched hardware., Trained what it describes as the first transcoders on a vision-language-action (VLA) model to predict robot actions before they occur.

## Products & services

- **Autonomous wash-and-fold laundromat** — Parametric's first autonomous physical business: dirty laundry in, folded clean laundry out, handled by robots operating inside a real wash-and-fold operation in San Francisco.
- **Bimanual mobile robot** — A wheeled robot with two arms, designed for long uptime and high throughput, that sorts, handles, and folds laundry.
- **RL feedback pipeline (RLHF for robots)** — An automated pipeline that combines customer feedback with a judge model to reinforcement-learning fine-tune the robot's policy, steering behavior toward what customers define as good - letting robots learn new tasks on-site with under an hour of data.

## Achievements

- Reported 3x higher task reliability than a state-of-the-art baseline model on matched hardware and benchmarks.
- Reported robots beating Physical Intelligence's Pi0.5 model with 20% greater speed on matched hardware.
- Trained what it describes as the first transcoders on a vision-language-action (VLA) model to predict robot actions before they occur.
- Built an RLHF-style feedback pipeline for robotics, similar to those used by frontier language labs.
- Deployed robots into a real, high-volume wash-and-fold operation in San Francisco.
- Accepted into Y Combinator's Fall 2025 batch.

## Latest updates

- **2025-12** — Launched publicly via Y Combinator, showing robots that learn to fold laundry from customer feedback inside a San Francisco wash-and-fold operation.
- **2025-11** — Y Combinator highlighted Parametric as a robotics company using RL to automate repetitive physical work.

## Links

- Website: https://parametric.company
- LinkedIn: https://www.linkedin.com/company/parametric-pbc
- Twitter/X: https://twitter.com/parametricpbc
- YouTube: https://youtu.be/EhiYJT52EJY

---

Profile page: https://yespress.io/parametric-yc-f25
Published by YesPress — https://yespress.io
Last updated: 2026-07-30