Day 5 of Build in Public & the Rear Casing for my Jet 🛩️ Engine is finally done! ✨🚀🛰️🌊☄️ @Siemens Solid Edge
We build physics-informed ML you can drop straight into real control loops, world models for touch in fluids & hardware.
An open-source inverse PINN engine that takes governing equations and noisy sensor data, discovers unknown physical parameters in real time, and outputs physics-guaranteed dynamics models that drop straight into MPC, RL, and state-estimation pipelines.
A coordination platform built on Hierarchical Federated Multi-Agent RL with PINN integration. It enforces physical limits with embedded electrochemical and thermal models, estimates hidden asset health through inverse physics, and preserves data privacy — augmenting existing VPP infrastructure rather than replacing it.
A purely tactile world model for soft-robotic manipulation in fluids — no cameras, no gel sensors. Perception comes from distributed strain along a tendon-driven Cosserat rod manipulator, where the deforming body itself is the sensor. A PINN embedding the rod equations and slender-body hydrodynamics generates dense synthetic tactile fields, so the model can plan through predicted touch where vision fails — like underwater coral manipulation.
Follow the research, ship a PR, or just argue about loss functions in the community.
New papers, releases, and milestones. No noise.
Day 5 of Build in Public & the Rear Casing for my Jet 🛩️ Engine is finally done! ✨🚀🛰️🌊☄️ @Siemens Solid Edge
First Building Block of my soon to-be Robot has arrived 🪸
NVIDIA launched Cosmos 3: world models as infrastructure for physical AI 🌊 But it's vision-first. Warehouses, roads, tabletops. What about where cameras fail? Underwater. Inside the body. Soft robots handling fish, threading vessels, gripping subsea equipment 🐙
Biomimetic robotics is the future >>>