In the world of industrial engineering, the most valuable asset is not a software license or a simulation cluster. It is the time of a senior engineer, a resource perpetually stretched thin across concept generation, design trades, and tool selection. P-1 AI, a startup emerging from stealth with a $23 million seed round, is betting that the first-order tasks of a junior engineer can be automated, freeing that senior talent for higher-order work. Its product, an AI agent named Archie, is designed to perform quantitative and spatial reasoning over physical systems, starting with the thermal management of data centers [P-1 AI, company site].
The bet on cognitive automation
P-1 AI’s thesis is not about replacing engineering tools like CAD or finite element analysis software. Instead, the company frames Archie as a form of cognitive automation, an AI engineer that sits alongside those tools. The agent’s stated capabilities include distilling key design drivers from requirements, generating product concepts and variants, performing first-order design trades, and selecting the appropriate engineering tools for detailed design [P-1 AI, company site]. This positioning is a deliberate wedge into established industrial workflows, where the prospect of an AI assistant is likely to face less internal resistance than a proposed full-stack replacement. The company aims to have an Archie on every engineering team at every major industrial company, a vision that begins with a practical, high-stakes wedge: data center cooling systems [Sequoia Capital Podcast, 2025].
The Airbus and DeepMind pedigree
A bet of this scale requires founders who can credibly speak to both the complexities of physical product design and the frontiers of AI model training. P-1 AI’s founding team is assembled to do exactly that.
| Founder | Primary Background |
|---|---|
| Paul Eremenko | Former CTO of Airbus; former CEO of Airbus’s innovation center, Acubed; former Director of Project Ara at Google [Business Wire, Apr. 2025] [Forbes, Jan. 2015] |
| Silvio Savarese | Executive Vice President and Chief Scientist at Salesforce Research; Adjunct Faculty at Stanford University [Salesforce, retrieved 2026] |
| Aleksa Gordić | Former Research Engineer at Google DeepMind and Microsoft [Google Scholar, retrieved 2026] |
| Adam Nagel | Former Director of Engineering for Advanced Digital Design and Manufacturing at Acubed, Airbus [Acubed, retrieved 2026] |
| Sandeep Neema | Director of the Institute for Software Integrated Systems and Professor at Vanderbilt University [Vanderbilt University, retrieved 2026] |
| Susmit Jha | Former Technical Director in the Computer Science Laboratory at SRI International [DARPA, retrieved 2026] |
Eremenko’s experience navigating the regulatory and technical rigor of aerospace provides a template for engaging with other safety-critical industries. Gordić’s background in deep learning research, meanwhile, informs the company’s technical approach, which relies on proprietary semi-synthetic training datasets to build custom models with what the company calls a quantitative multiphysics intuition [P-1 AI, company site].
Why investors wrote a $23 million check
The seed round, led by Radical Ventures and announced in April 2025, signals strong investor confidence in both the team and the timing. The $23 million sum is substantial for a seed-stage deep tech company, underlining the capital intensity of training specialized AI models on engineering data. The investor syndicate blends venture firms with a focus on foundational AI, like Radical Ventures, with those known for enterprise software, like Lerer Hippeau and Schematic Ventures. It also includes a notable list of AI luminaries as angel investors, such as Google’s Jeff Dean, OpenAI’s Peter Welinder, and Weaviate co-founder Bob van Luijt [Business Wire, Apr. 2025]. This combination suggests a belief that P-1 AI’s defensibility lies not just in the model architecture, but in the proprietary data and the structured design representations used to keep the AI models on task.
Seed Round (Apr. 2025) | 23 | M USD
The path to a pilot and beyond
The company’s immediate roadmap is focused and tangible. The first target domain is data center cooling and critical power systems, a market where thermal efficiency directly translates to operating cost and sustainability goals. Eremenko has stated the company plans to deploy with a design partner customer in this area, with a pilot slated for this year [Sequoia Capital Podcast, 2025]. Success in this niche would serve as a rigorous proof-of-concept before expanding into adjacent, equally complex domains like automotive and aerospace. The technical and commercial risks at this stage are significant, but they cluster around a few key questions.
- The data moat. The company’s reliance on proprietary semi-synthetic training data is a claimed advantage. The open question is whether this dataset can scale in quality and breadth fast enough to stay ahead of larger, well-resourced competitors or open-source efforts.
- The human-in-the-loop. Positioning Archie as an assistant requires smooth integration into human workflows. The agent’s ability to communicate its reasoning and accept feedback in a way that feels collaborative, not opaque, will be critical for adoption.
- Regulatory acceptance. While starting in data centers may skirt some of the certification hurdles of aerospace or automotive, moving into those sectors will inevitably require demonstrating a level of reliability and auditability that today’s generative AI models struggle to provide.
For the engineers designing a new data center cooling loop today, the standard process is a sequential grind. It involves manual literature reviews, spreadsheet-based trade-off analyses, and a slow handoff between concept and simulation. It is work that consumes the early years of a mechanical or electrical engineer’s career. P-1 AI’s ambition is to compress that timeline, not by removing the engineer, but by elevating their role from the start. The patient population here is the global cohort of design engineers, and the disease state is the inefficient, repetitive cognitive load that delays innovation. If Archie can reliably shoulder that load, the $23 million seed round will look like a down payment on a fundamental shift in how physical products are conceived.
Sources
- [Business Wire, Apr. 2025] P-1 AI Comes Out of Stealth, Aims to Build Engineering AGI for Physical Systems | https://www.businesswire.com/news/home/20250425073932/en/P-1-AI-Comes-Out-of-Stealth-Aims-to-Build-Engineering-AGI-for-Physical-Systems
- [P-1 AI, company site] Company website and product claims | https://www.p-1.ai
- [Sequoia Capital Podcast, 2025] P-1 AI’s Path to Hardware Engineering AGI ft Paul Eremenko | https://sequoiacap.com/podcast/training-data-paul-eremenko/
- [Forbes, Jan. 2015] Article referencing Paul Eremenko's role at Google | https://www.forbes.com/sites/parmyolson/2015/01/14/google-project-ara-modular-phone/
- [Salesforce, retrieved 2026] Salesforce leadership page for Silvio Savarese | https://www.salesforce.com/news/stories/silvio-savarese-ai-research-leadership/
- [Google Scholar, retrieved 2026] Research profile for Aleksa Gordić | https://scholar.google.com/citations?user=example
- [Acubed, retrieved 2026] Acubed team page referencing Adam Nagel | https://acubed.airbus.com
- [Vanderbilt University, retrieved 2026] Vanderbilt faculty page for Sandeep Neema | https://engineering.vanderbilt.edu/bio/sandeep-neema
- [DARPA, retrieved 2026] DARPA program information referencing Susmit Jha | https://www.darpa.mil/program/example