Foundation EGI, an MIT spinoff, emerges from stealth with $7.6M seed round backed by Samsung & Henry Ford Health

Foundation EGI, an AI engineering startup spun out of MIT, has officially launched from stealth with a $7.6 million seed round. The funding was led by E14 Fund and Union Lab Ventures, with participation from Samsung Next, GRIDS Capital, Stata Venture Partners, and Henry Ford III.

The company is taking on one of the most overlooked pain points in manufacturing: messy, manual engineering processes that haven’t kept up with the digital transformation seen in other industries.

Engineering General Intelligence (EGI)

Foundation EGI says it has built the first domain-specific AI platform focused on engineering general intelligence (EGI). The platform is already in testing with Fortune 500 industrial brands, where it’s being used to clean up the chaos in product design and development.

The team isn’t short on pedigree. The company was co-founded by MIT researchers Mok Oh, Professor Wojciech Matusik, and Michael Foshey. They’ve put together a group of engineers and operators with deep experience in manufacturing, AI, and enterprise software.

The pitch is simple: turn vague, disorganized engineering instructions into clean, structured, machine-readable code. The platform is web-based and integrates with existing design and manufacturing software. Foundation EGI’s custom-built large language model is trained specifically for engineering and understands physics, spatial logic, and how physical products actually get made.

“Engineering is primed for an AI shift, but generic LLMs aren’t built for this world,” said CEO Mok Oh. “We’ve built a domain-specific AI that transforms messy documentation into structured, accurate processes, starting with engineering documentation. The result: faster product cycles, fewer mistakes, and serious cost savings.”

Why Now?

Unlike other industries that have embraced digital transformation, manufacturing and engineering remain stuck in outdated workflows. Instructions are often scattered, handwritten, or buried in tribal knowledge, slowing teams down, introducing errors, and contributing to what Foundation EGI estimates is more than $8 trillion in annual economic waste.

That’s the gap Foundation EGI is aiming to close.

The MIT spinoff has built a domain-specific AI platform that turns messy, inconsistent engineering instructions into structured, machine-readable code, bringing clarity, speed, and accuracy to every stage of product development.

At the core is a purpose-built large language model trained specifically for engineering. Combined with its agentic AI platform, Foundation EGI takes natural language inputs—no matter how vague or unstructured—and converts them into accurate, codified instructions. The web-based platform integrates directly with the major design and manufacturing tools already used by engineering teams, making adoption seamless.

By transforming tribal knowledge into structured systems, Foundation EGI brings automation and transparency to workflows that have historically been disorganized and inefficient, helping teams move faster, reduce delays, and build better products.

Click to see larger image

Early users like Inteva Products, a global automotive supplier, are already on board. “It’s clear Foundation EGI will help us eliminate unnecessary costs and automate disorganized processes,” said Dennis Hodges, Inteva’s CIO.

Investors say the market timing is spot on. “Engineering and manufacturing have been begging for this kind of AI solution,” said Habib Haddad, founding partner at E14 Fund. “The team, the tech, and the market conditions make this a rare opportunity to tackle a long-standing industrial bottleneck.”

Co-founder Professor Wojciech Matusik also shared the company’s vision on stage at TEDx MIT, describing how EGI turns natural language into engineering-specific instructions grounded in real-world physics. “This unlocks creativity and speed for a new generation of engineers,” he said.

The startup’s foundation is built on years of research at MIT and other academic institutions. That work culminated in a March 2024 paper titled Large Language Models for Design and Manufacturing, co-authored by Matusik, Foshey, and others.