![]()
Dr. Valerio Tenace is CTO and Co-Founder of ChipNexus, developer of agentic AI software for semiconductor and hardware design, and an expert in hardware design automation and machine learning applications for electronic systems. In his role at ChipNexus, Valerio drives the technical vision and architecture behind its NEX AI platform.
What brought you to ChipNexus?
ChipNexus grew out of a problem I’d been working on in academia and industry. In late 2021, I began exploring how to make it easier for engineers to go from a specification written in natural language to synthesizable RTL. That work led us to found PrimisAI, which is now ChipNexus. Generating RTL was our first milestone, and it soon opened up a bigger question: how much of the silicon design flow could we automate? Today, our solutions cover the entire design stack.
What drew me in from the beginning was the chance to make engineers more efficient as they take a design from an idea to silicon.
What is the biggest challenge your customer is facing?
Time to market. A chip design project takes a lot of engineering effort. Teams have to define specifications, keep documentation in sync with the design, explore architectural options, and verify the results at every stage. More often than not, they also have to migrate an existing design to a new process node or technology. Needless to say, changes in one part of the design flow often trigger rework elsewhere. That’s why we want to optimize the full design cycle.
We look at where teams spend their time across the project, including on repetitive tasks. If we can help them get through that work faster, they can lower development costs and spend more time solving higher-level challenges.
Why is your approach to solving this problem unique?
In late 2023, while we were still working on our front-end design product line, we started building Nexus: an open-source platform and one of the very early multi-agent systems to use large language models for hardware design. We gave it a hierarchical structure so it can break a complex engineering task into smaller units of work and route each one to the agent best suited to handle it.
Since then, we’ve improved how the system follows the engineer’s instructions as a task evolves. A hardware design task often takes several iterations. The engineer might refine a requirement or redirect the work based on an intermediate result, and the system needs to follow that guidance. We’ve designed the interaction around the engineer, who sets the direction and evaluates the result.
More recently, we’ve been developing proprietary AI models in-house to further speed up some of the most time-consuming tasks in the design flow. We’ll share more about those models in the coming weeks.
What other approaches are being used to solve this problem?
General-purpose AI assistants are one approach. They can help with tasks like drafting documentation and generating code. However, reliability and trustworthiness are key when AI is used in silicon design: an error introduced early can propagate downstream, where it becomes harder and more expensive to fix. An assistant also needs precise knowledge of the EDA tools, IP blocks, and design guidelines a team uses.
Giving a language model the right context at each step takes careful work. Also, most chip companies need to keep their design IP within their own environments. For those teams, local execution is essential and that alone brings its own engineering challenges.
What challenges will your customers face in the near future?
I think one of the biggest challenges will be learning how to align these systems with design goals and steer them as the work progresses. That makes a deep understanding of hardware design more important than ever. When an assistant proposes an architecture, for example, the engineer needs to spot an assumption that could cause problems later, even if the proposal looks sound on paper.
In some ways, I see the hardware engineer taking on the role of a supervisor. They’ll define the problem, set the constraints, and give the assistant feedback as it works through the task. That expertise matters whether the assistant is helping explore a new architecture or taking on repetitive verification work.
How has AI changed your business strategy?
AI has been the cornerstone of our strategy since we started. The range of problems we can tackle has grown, and what we learn from customers helps us decide where to focus next. I’m particularly interested in how engineers use our tools in their day-to-day work. I want to know which tasks they keep coming back to, when they step in to guide the system, and what new uses they find for it.
Those conversations show us where our tools are already saving engineers time and where the next improvements could make the biggest difference. That’s what I’m most excited to build on.
Also Read:
ChipNexus Brings Agentic AI into Established Analog Design Flows
Share this post via:

Comments
There are no comments yet.
You must register or log in to view/post comments.