Home / Technology / Agentrys: AI That Learns Your Chip Workflows
Agentrys: AI That Learns Your Chip Workflows
4 Aug
Summary
- AI alone may not create lasting chip design advantage.
- Agentrys platform combines workflow knowledge and learning.
- Its approach builds customer-specific, compounding engineering capability.
Mark Ren, formerly of Nvidia, has launched Agentrys to provide a distinct approach to AI in chip design. He posits that while frontier AI can automate tasks, it won't create lasting competitive advantages if all companies access similar tools. Agentrys's platform emphasizes integrating customer-specific workflow knowledge with continuous learning through agent-native tools.
This strategy aims to build a proprietary engineering capability that compounds over time. Agentrys's 'Agentic Design Automation' combines AI agents, engineering knowledge, and learning to support complex chip-design workflows. The platform operates within customer environments, ensuring sensitive data remains secure and the AI workforce learns from their specific designs and methodologies.
The Agentrys approach focuses on automating entire workflows, not just isolated tasks. The 'Self-Evolve' framework captures observations from each workflow execution, refining agents and knowledge based on customer-defined criteria. This aims to create a continuously improving agentic engineering workforce unique to each semiconductor company.
Agentrys counters traditional EDA vendors by offering a cross-ecosystem solution that accrues benefits to the customer. It addresses data thinness concerns by leveraging repetition and structured feedback from agent-native tools. The goal is to build a defensible advantage based on an organization's accumulated design decisions and methods, rather than easily replicated task-level AI.
