Artificial intelligence-native cloud infrastructure giant CoreWeave Inc. says it’s going to help enterprise engineering teams to implement AI directly into their workflows through its new Physical AI Field Engineering service.
Announced today, the new service is meant to bridge the massive gap between industrial domain expertise and applied machine learning. CoreWeave has recruited a team of specialized engineers with expertise in industries such as the automotive, aerospace and mechanical engineering sectors, who will work directly with customer’s own engineering teams. Together, they’ll collaborate on AI projects, training models on the customer’s data and integrating them with their workflows and applications to ensure their investments in the technology pay off.
CoreWeave is tackling a critical talent gap that exists in almost every industry. For instance, in the aerospace sector, most companies have plenty of domain engineers with immense knowledge of complex physical systems, such as aerospace structural loads and combustion dynamics.
They may also have a team of developers with experience in developing AI models. But what they don’t have is AI developers who are knowledgeable enough to work with their domain engineers, because they lack a basic understanding of the physics of the systems they’re trying to integrate AI with.
This is where CoreWeave’s teams of specialized engineers can make a difference. By working closely with customer’s own teams, they can help to make sure that whatever AI models they develop hold up against the physical realities of the systems they’re designed to optimize.
CoreWeave said each new Physical AI Field Engineering engagement begins with a workshop, where its engineers will work with the customer’s chosen team to evaluate their engineering workflows, identify the best use cases for AI, and then establish a quantified return on investment before making any major commitments.
From there, the next step is to design and build the actual models using the customer’s existing and real-time data to try and predict physical outcomes. The data from that analysis can then be used to identify which information should be used next to further inform model development, and in that way, CoreWeave says, it can cut testing times by between 17% and 35%.
After this comes the infrastructure, which is where CoreWeave’s core expertise comes into play. It will help customers to set up the most optimal compute environment for their AI projects, taking care to avoid over- or under-provisioning resources so they get the best value for money. Once things are up and running, the agentic learning process begins – where AI insights are transformed into physical actions, so that robots can execute trained skills, systems can correct faults in machinery before the equipment fails, and so on.
The final step involves integrating the AI into customer’s existing workflows — in the shape of working applications, dashboards and optimization tools — so they can immediately implement their new capabilities and see the benefits.
CoreWeave said the entire Physical AI Field Engineering service is underpinned by its cloud infrastructure, which includes its specialized bare-metal servers and a suite of integrated engineering tools. These include Weights & Biases’ AI evaluation platform Weave and its specialized models for experiment tracking, the marimo tool for data exploration and ARIA for developing autonomous agents.
Senior Vice President of Physical AI Richard Ahlfeld said it takes more than a fancy demo to convince engineers to adopt new tools and methods. “They adopt it after it has held up in their own hands, on their own systems,” he said. “That is why we send engineers who speak the same language as the teams across the table, and why we build on the customer’s own data instead of handing back a report someone else needs to implement.”
The company reckons it has already had more than 100 engagements with early adopters in the automotive, aerospace and robotics industries. One of the first was Nissan Motor Co., which CoreWeave helped to create new predictive models based on 90 years’ worth of previously untapped archived test data to optimize chassis bolt-joint evaluations, reducing physical testing times by 17%. At a second, unnamed automaker, CoreWeave’s engineers used its proprietary data to complete a key engine calibration step, which normally took three months, in just 24 hours.
CoreWeave says the new Physical AI Field Engineering service stems from its acquisition of a startup called Monolith AI Ltd. last September, which pioneered the use of AI and machine learning to solve complex challenges in physics and engineering.
