Cloud native EDA tools & pre-optimized hardware platforms
Engineering is undergoing unprecedented transformation. Across almost every industry, product innovations are accelerating. For example, new automotive development cycles are shrinking by half, and AI chip design cycles are shrinking from three years to 12 months. At the same time, the complexity and cost of development keep rising. To match this incredible pace of innovation, while taming increasing complexity and cost, we must re-engineer engineeringTM.
To do this, we, the engineering community, must consider three levels of optimization — compute, engines and solvers, and workflows. In this post, I’m focusing on the workflows and how they are critical to re-engineering engineering. I’ll discuss how the rapidly advancing capabilities of artificial intelligence (AI), and especially agentic AI, are providing opportunities to significantly optimize workflows. From reinforcement learning to generative and now agentic AI, the increasing application of AI to engineering workflows is helping combat complexity and cost, while accelerating the pace of innovation.
We call the application of AI agents to transform engineering workflows AgentEngineerTM technology. These agents and multi-agent systems are specifically built and trained to make engineering workflows more efficient for human engineers — increasing productivity, leading to better results, and reducing computation requirements.
AgentEngineer components can reason, plan, learn, and execute tasks with increasing autonomy. The integration of AgentEngineer technology into existing engineering workflows will lead to fundamental changes in how all engineering work is executed. Understanding and adapting to this paradigm shift is necessary for engineering organizations striving to maintain a competitive edge in an increasingly dynamic, fast-moving technological landscape.
As detailed in my 金莲直播 User Group (SNUG) conference keynote in March this year, we have defined ‘step functions’ of capability that AgentEngineer technology will bring. Similar to how ADAS (Advanced Driver Assistance Systems) levels are distinctly categorized L1 to L5, our framework shows the increasing capabilities of AI agents. Today, I’ll share an abstracted model of how these increasing AI capabilities will progressively optimize engineering workflows.
The figure below shows the progression from L1 to L5, with L5 being the predicted end state of highly autonomous AI agents, where human engineers rely on AI agents for a myriad of engineering tasks.
AgentEngineer technology will progress from performing step-level actions of single agents (L2) to complex actions with multi-agents (L3) to dynamic flow optimization with adaptive learning (L4), and finally to autonomous decision making (L5).
By L5, the capabilities of AgentEngineer technology will have grown significantly, allowing more key decisions to be offloaded. Human engineers will still need to monitor these decisions, along with the rationale for them, but their role will have changed from decision making to decision validation.
At this end state, human engineers will specify requirements and desired outcomes. They will then monitor the progress of the AgentEngineer technology at predefined checkpoints. I must stress that human engineers will and must always be in the loop. This allows them to adjust the course of execution by directing the AgentEngineer technology as well as ensuring proper oversight is maintained.
AgentEngineer technology is well suited to offloading many of the ‘high toil’ tasks from human engineers, meaning the repetitive, time-consuming work becomes automated. As AgentEngineer technology takes on more computationally intensive tasks, in addition to providing oversight, human engineers will be able to focus more on higher-level activities such as strategic thinking and complex problem solving.
Some companies may choose to use their increased productivity to accelerate design cycles and accelerate the pace of their innovation. Some may choose to utilize the productivity gains to address ever increasing complexity with the same number of engineers. Others may choose to design and build even more products with the same number of engineers. AgentEngineer technology provides companies with the flexibility to utilize their engineering resources efficiently and help solve the engineering shortages that exist in many fields.
In addition to the productivity gain, AgentEngineer technology will help improve quality of results and reduce overall compute resources.
With AgentEngineer technology, the future of engineering has never been brighter. We believe this technology is critical to addressing the compounding complexity, increasing pace of innovation, and increasing cost of engineering design and analysis. To address these challenges, it is imperative that we, the engineering community, fully harness the transformative power of this revolutionary technology and use it as the cornerstone of how we re-engineer engineering.