A few colleagues from MIT have raised significant capital for a startup focused on “Scientific ML” in chip design and hardware development. They are now hiring folks with experience in anything related to reverse design, ODE and PDE autodiff, neural ODEs, physics-inspired neural nets, optimization, classic ML, LLMs, design automation, and the creation of automated harnesses around industry-standard EDA and multiphysics design tools.
You can see the open positions at https://www.axiomatic-ai.com/
It is a pretty good opportunity in terms of joining early and gaining equity in the startup. They sponsor visas and have options for on-site and remote positions.
11 Likes
From conversation with these folks, it seems they plan to have the glue/core of their code base in python (as a glue language and as a harness for industry-standard EDA tool automation), but to lean mostly on Julia and Python/Jax for the reverse design, simulation, optimization, autodiff, etc, mostly depending on what their hires prefer.
If you are a Julia hacker, you probably will have the freedom (and be encouraged) to build in Julia.
1 Like
Hi Dr. Krastanov,
I’m Vivek, an Algorithm Developer for Physics and Engineering applications with an interest and experience in Neural Networks, Mathematical Modeling and Signal Processing. I also like Sports - I run Trail Ultramarathons and have a M.Sc. Sports Engineering Degree in addition to a background in Physics (BS-MS Physics).
am currently working on the development of an architecture for a Physics-informed Neural Network (PINN) directed towards solving an inverse problem in Optics (scattering-related) for a “data-scarce” setting. The specific technique that I’m exploring is a Neural Radiance Field (NeRF) inspired Convolutional Neural Network (CNN) coupled with “unconventional” Data Augmentation. The architecture is being developed in Python via JAX and PyTorch.
I have also taught PhD level classes on Physics-informed Neural Networks - designed PyTorch tutorials on ODE and PDE autodiff, physics-inspired neural nets and classic ML regression. In addition, my previous research project involved optimization and was focused on the development and implementation of algorithms for motion tracking from noisy sensor data.
My primary experience is with Python. However, at least for wave-based inverse-problems in Optics, there’s scope to develop a " Complex-valued" Physics-informed Neural Network package “ab-initio” since complex-valued perceptrons have greater functionality that their real counterparts and complex numbers are natural to wave-based Physics. Julia seems the way to do that which is why I’ve been thinking about Julia - which is why I was on the Julia website - and came across your post - to give you some context!
I would really like to know more about this exciting job opportunity and discuss things further. I have also emailed you but I don’t know if I sent it to the correct Dr. Krastanov - It was a UMass email. To be on the safe side, I decided to reach out here, on the UMass email and on the Axiomatic AI jobs email.
I’m hoping to hear back soon!
Thanks and Kind Regards,
Vivek
Hi, Vivek! Your background sounds quite appropriate for the postings. The jobs email is indeed the correct place to send your inquiry. I would expect them to respond to you soon, let me know if that is not the case.
Hi Dr. Krastanov, thanks for your prompt response and reassurance! The job sounds very exciting. I will definitely let you know if I don’t hear back from them soon.
You mentioned that they have options for on-site positions. Please could you elaborate a little on the on-site positions? Where is the physical office for the company?
It would be best to consult the job postings at their website. The job listings seem to show Boston as one of the main locations.
Thank you! I expected it to be near MIT. Have a great weekend and I’ll let you know in case I don’t hear from them by Tuesday 9th April 2024.
Dear Dr. Krastanov, I heard back from the company and booked a meeting with their CEO Dirk Englund.
Thank you very much for sharing this job posting!
1 Like