Founder-engineer building AI systems for the physical world. PromptPCB, AutoTQ, robotics, PCB design, firmware, ML, and medical devices.
Joined July 2019
- Tweets 305
- Following 87
- Followers 79
- Likes 68
113 Photos and videos
Pinned Tweet
May 31
I’m rebuilding this account around what I actually do now:
I build AI robotics systems for the physical world.
PCB design. Embedded C. Robotics control. ML/computer vision. Mobile apps. Medical-device engineering. Mechanical prototypes. The messy stuff where software has to survive contact with hardware.
I’ve founded and built projects across pathology imaging, automated agriculture, medical devices, AI training tools, PCB automation, and robotics. I’m currently building PromptPCB and leading engineering on AutoTQ.
This account is going to be build logs, technical notes, failures, fixes, prototypes, AI hardware workflows, and lessons learned the hard way.
No guru content. No fake founder theater.
Just engineering, product building, and the grind of making difficult systems work.
4
385
Devin Willis retweeted
Jun 11
The way I would automate PCB design is spending a year or so doing electrical engineering classes to learn how to make a simulator and then training models in those simulators as verifiers
Jun 11
This animation shows the electric field distribution inside the shielding can of the framework SDR module, with one of the traces acting as the excitation source. I need to run the same simulation for the filter path, but it looks like I'll need absorbers in the cavity
26
16
359
30,030
Jun 11
anyone know of any projects doigngLLM memory management inspired by os memory paging techniques
but maybe semantically aware
seems like an extremely similiar problem
1
26
Jun 10
If you move past using google flow to make clips with people or animals where its squarely in uncanny valley
its actually insane for general B-Roll, promo video, static shots
22
Devin Willis retweeted
Jun 9
Holy shit it can answer the strawberry question
387
369
10,144
965,321
Jun 8
current pulse on AI PCB design seems like its still in a stage where it lacks grounding and is prone to hallucination or misses complex interactions and behaviors
i'm confident this is a solvable problem
I've seen some thoughts that there isn't enough data or there wasn't enough training data in this domain
but i think all the data thats needed is present in datasheets and in simulation data that can be gathered from the circuit and it just needs to be collected and structured in a way that allows for AI based reasoning
for humans that involves reading datasheets and looking at schematic files, and drc tools, etc
for AI to be capable I think the input data needs to go beyond dropping in a netlist, or schematic, or footprints etc and needs to use deterministic tools in order to help pull out signal from noise on behalf of the AI in order to unlock its reasoning ability in a grounded way
1
28
Jun 7
still unreal that I got the chance to briefly talk about the work i'm doing at app.promptpcb.ai with @alexwg thanks to @EwillieP for making the whole thing happen
youtu.be/wR9-wtAN2Bw?si=S7aV…
2
86
Devin Willis retweeted
New interview with the University of Florida AI² Student Club is out.
Jun 7
Every sci-fi reality will arrive in the next 10 years.
Less than a week ago I cold DM'ed frontier physicist Dr. @alexwg for a UF AI club Q&A.
He replied in minutes, said "email me," and we were live 3 days later.
That speed says everything—about the urgency of this moment and who Alex is.
Our full 2-hour conversation on agency, physics, and the next decade: youtu.be/wR9-wtAN2Bw
11
10
101
12,295
Devin Willis retweeted
Jun 6
we're going to re-invent PCB manufacturing and win
4
2
49
1,709
Jun 6
any opinions/advice on setting up your own pcba line
we have plenty of space and power
1
55
Jun 5
We're so back
Congrats to @lennykhazan! 👑
Let's keep the party going with a new prize.
$1000 for best score by 6/21. Must be >22.
1
121
Jun 5
how is this a real number
Jun 5
JUST IN: Elon Musk's SpaceX signs $920,000,000 per month deal with Google to provide AI computing power.
64
Jun 5
restarting on world models to prep for more robotics. started as simple as it gets: a 1D world of bouncing balls, and a net that only sees positions.
it predicts the future, recovered velocity conserved energy on its own…
and you can reach into its latent, edit a ball's "velocity," and watch its future bend.
1
1
68
Jun 5
world models are the core bet for robotics and i haven't really started with them from first principles
so I shrank one until I can see everything. 1D world, 3 balls, known physics. fed only positions, the net forecasts the future and builds its own velocity conserved energy internally (R²=0.94 each, ~0 from the raw input ).
I can edit those latents and the future changes accoridngly (slope 1.00).
next will be a larger 2D world model with richer physics, under the same microscope
i'm going to work my way up to something usable
37
Jun 5
looking for people building hardware for new startups
tag them here
7
13
757