What are you world-class at? What are you super competitive about?

Brad: Building compelling products and product teams. At ApplyBoard, working directly under the CEO, Martin Basiri, I wrote the white paper on ApplyX: the foundations for how ApplyBoard approaches innovation and new product lines. Within ApplyX, I led the early product discovery for a product line that is now doing nearly $100M in revenue and led the discovery and launch of 2 other product lines.

Zeyi: Top 100 in the world for deep reinforcement learning and planning algorithms. He came from the research group that laid the foundations for AlphaGo. For his thesis project, he developed a variant of MuZero, the most recent version of AlphaGo. His work is regarded by experts as the best publicly available version of MuZero implementation. Outside of DeepMind, Zeyi is one of the few people with both the software engineering skills and AI knowledge to build a full-fledged MuZero system.

Relevant context about MuZero, AlphaGo and DeepMind:

• Much of the foundational AI breakthroughs for AlphaGo came from the research group of Martin Muller.
• The lead researcher for DeepMind’s AlphaGo, David Silver, came from Martin Muller’s research group at the University of Alberta. The same group Zeyi was a part of.
• AlphaGo can only play Go, but two years ago DeepMind published MuZero, a more general decision-making AI that can achieve superhuman-level performance in a broader range of applications.
• Only DeepMind has access to the source code of MuZero, so anyone else who wants to utilize this state-of-the-art decision-making AI has to implement their own from scratch.

Why are you choosing to do this over another business?

Brad: I have had a passion for physics and electrical engineering, and I have been tinkering with and building electronics ever since I can remember. On a higher level, Improving global computation is incredibly beneficial for the world, especially now in the age of increasingly large AI models. Reducing the overblown cost of microchip design will bring huge advancement in this space.

Zeyi: Astrus is the perfect blend of my passions for cutting-edge technology in AI and semiconductors. The opportunity to work on an unsolved problem using the most advanced technology available in both fields, while making an impact on the world, is truly unique and compelling. It is an opportunity for me to not only advance my knowledge but also make a real difference.

Google published a paper on using deep reinforcement learning for chip design and have leveraged the tech internally - where do you see the need for further validation, and where do you see the opportunity for your company given existing research/work in the space?

The Google paper addresses the problem in the digital domain, while Astrus focuses on solving a similar problem in the analog domain. The digital place and route problem is already automated in existing workflows, and even if Google productize their algorithm, it will only be a 10% improvement over the existing alternative. Analog layout, however, is much harder, still completely manual, and represents over 75% of the design effort.