Wu Xing

Founder of Simulation × AI Surrogates. Neural operators.Turning thermal margin into performance.

[Wu Xing @Zen_Tea_Exp]

@rachell331 Welcome, Rachel! 👋 Engineer here, using AI to replace slow physics simulations with fast neural surrogates.

[Wu Xing @Zen_Tea_Exp]

One NVIDIA H100 GPU puts 700 W into a liquid-cooled cold plate. We built a 3-D thermal simulation on ThermaLoop (MIT License), then trained an AI surrogate model on it: the same temperature map 17,473× faster, within 0.5 K. Simulations stop being one-off reports and become a daily tool: 100,000 cooling designs screened in 2.6 s instead of 13 hours, and, when a pump weakens, a live hotspot map from three sensors plus the safe coolant flow.
#NVIDIA #H100 #GPU #AI #SurrogateModel #LiquidCooling #ThermalSimulation

[Wu Xing @Zen_Tea_Exp]

NVIDIA PhysicsNeMo's official example: conjugate heat transfer on an industrial heat sink, no mesh.

How it works: Navier–Stokes and conduction equations become the loss function. Four networks learn flow, temperature and heat flux, coupled at the fluid–solid boundary.

#CFD #PINN #PhysicsNeMo

[Wu Xing @Zen_Tea_Exp]

I used Claude to rebuild NVIDIA PhysicsNeMo's industrial heat sink example: 44 copper fins, conjugate heat transfer, no mesh, four neural networks.

Their result holds up:
Peak temperature within 4 °C of CFD.
Pressure drop 9–12% high.

The second line is the one engineers should read. A PINN nails the number you check first and drifts on the one you don't.

Transfer learning makes new designs 5× cheaper. That's only worth it if you know where the model is wrong.

[Wu Xing @Zen_Tea_Exp]

After compute, the next wall for humanoid robots is heat.

A joint winding hits 105°C in an hour of continuous work — then the motor derates.

60 seconds on how robots actually stay cool 👇

[Wu Xing @Zen_Tea_Exp]

Trained a DeepONet on it: one network learns the spatial modes, another learns how the operating condition weights them. 0.35 °C MAE on conditions it had never seen, versus 3.1 °C from the old 8-point model.

#CFD #MachineLearning #ThermalDesign #NeuralOperators

[Wu Xing @Zen_Tea_Exp]

@AppLauncher_App It's an interactive CNN visualizer — draw a digit and watch it flow through every layer, edit kernels, see why it predicted that class. Built it because formulas never gave me a mental picture.

https://t.co/cHol8W1Pe8

[Wu Xing @Zen_Tea_Exp]

I asked Claude to make a video for my app, and it just shipped this
https://t.co/cHol8W1hoA

[Wu Xing @Zen_Tea_Exp]

The useful question isn't "can AI do my task?" but "how much of my job needs me present?" High substitution with low complement is the red zone. Adding complement — judgment, accountability, physical presence — is what moves you right.

Chart: https://t.co/d54Oqmbp1x

[Wu Xing @Zen_Tea_Exp]

DeepONet: instead of learning one solution, it learns the operator itself. Branch net encodes the input function, trunk net encodes the query location, and their dot product gives the field at any point. Train once, then query any coordinate — no remeshing, no retraining.

[Wu Xing @Zen_Tea_Exp]

@MrLarus 不错

[Wu Xing @Zen_Tea_Exp]

@Gwynne_Shotwell The "Scotland" sticky notes were working overtime today.

[Wu Xing @Zen_Tea_Exp]

@demishassabis @elonmusk Why Demis

[Wu Xing @Zen_Tea_Exp]

In the middle of the picture of this work is a hut, surrounded by a few slender trees, the picture is light and alienated. I don't know why when I saw this work, It was reminiscent of the distant painter Gauguin. Perhaps both of them have a bumpy fate. Or the person himself...

[Wu Xing @Zen_Tea_Exp]

In 1276, Mongolia captured Hangzhou, the capital of the Southern Song Dynasty. Qian Xuan did not go to the Yuan Dynasty to be an official, but to live in seclusion. His green mountains and green waters reminds people of the gorgeous landscape paintings of the Tang Dynasty.

[Wu Xing @Zen_Tea_Exp]

"From a single leaf’s fall to Thoreau’s Walden, both reveal how nature’s smallest moments inspire life’s deepest truths."

[Wu Xing @Zen_Tea_Exp]

@ARTBYODC

[Wu Xing @Zen_Tea_Exp]

@ArtvisionNFT In the tender embrace of the earth, the green expanse whispers of simplicity. A lone farmer, bent in quiet labor, becomes one with the land, embodying the harmony we so often overlook. Here lies the essence of life—untainted, deliberate, and true.

[Wu Xing @Zen_Tea_Exp]

"I went to the woods because I wished to live deliberately, to front only the essential facts of life, and see if I could not learn what it had to teach, and not, when I came to die, discover that I had not lived."

#Walden

[Wu Xing @Zen_Tea_Exp]

@DarkDescentL The vibrant spirit of the little lives in autumn.
— Song Dynasty, Anonymous