Multiphysics Simulation for the AI era
Foundation models for physics simulations across electromagnetics, acoustics, fluidics, and more. We believe that now is the right time to take the technology for doing AI simulation out of the research lab and into practical application to real world industrial problems. Our objective is to make generalizable, fast, and accurate AI simulation available to engineers solving problems across a wide variety of different domains.
Integrated silicon photonics simulation using the effective index method. FDFD (left) vs. zero-shot result from our pre-trained AI model (right). The AI model was trained on 20k sample devices with a larger characteristic length scale. This demonstrates the generalizability of the AI model.
FDFD
Traditional commercial simulation software relies on numerically solving complex partial differential equations, inherently limited in speed and scalability.
Kronos AI Model
Orders of magnitude faster and just as accurate: Our AI-powered simulations provide accurate results in significantly less time compared to traditional methods.
Our Value Propositions
Speed and Generalizability
Our highly generalizable models provide precise physics insights for diverse practical applications orders of magnitude faster than current solutions, drastically accelerating product development and validation workflows.
Multiscale and Heterogeneous Designs
We enable the design and analysis of larger, more complex spatial-temporal systems, such as simulating microwave and photonics components together in silicon photonics interconnects—essential for modern data centers—or creating millimeter-sized flat meta-lenses with nanoscale features.
Making Inverse Design Ubiquitous
We support the growing demand for inverse design. Currently, most design tasks rely on human intuition and PhD-level expertise. With the shift towards AI-driven automation in engineering, inverse design is becoming increasingly relevant – but it requires orders-of-magnitude more simulations. Our faster simulations will make inverse design more ubiquitous, unlocking the vast space of solutions beyond human intuition.
Increased Design Flexibility and Accessibility
The massive speedup of our simulations relative to traditional methods means that, rather than taking hours or days to run simulations in batch, our simulations can instead be dynamically interactive. This allows both easy design-space exploration as well as allowing domain experts, who typically would not also be experts in numerical methods, to easily incorporate simulations testing into their work. Making simple-to-use simulation tools much more easily accessible will create new groups of users who can benefit from but could not previously access the technology.
Orders of Magnitude Acceleration for Electromagnetic Simulation
We provide fully pre-trained plug-ins
for existing software packages or standalone design software.
for Optoelectronics
Optoelectronic and photonic circuits will determine the next stage of development in computation, telecommunication, the internet, and cryptography.
- We offer orders-of-magnitude acceleration in the design of optoelectronic circuits of arbitrary complexity, including light sources, waveguides, resonators, and detectors.
for Electronics
Modern electronics rely on the seamless integration of multiple chiplets into complex 3D architectures. Interconnects–the cross-talk between them and delay time–become crucial in such circumstances.
- We offer orders-of-magnitude acceleration in the complete design of 3D electronic chips for the most demanding electronic solutions.
for Telecommunication
Future 6G telecommunication will rely on the efficient real-time rerouting of signals. This requires fast electromagnetic simulations, which modern computers cannot handle.
- We offer an AI solution for fast computation in the most demanding and complex electromagnetic environments, dramatically accelerating internet and telecommunication speeds.