Paris-based startup Freeda has just raised €3.4 million in pre-seed funding, led by Frst and Brick & Mortar Ventures, to scale its AI-powered plan review platform across Europe and beyond.
Freeda’s premise is deceptively simple: construction teams spend weeks manually checking architectural plans for compliance with fire safety, accessibility, other local regulations or brand standards and still miss errors that trigger delays and cost overruns.
With Freeda, a project owner can upload plans and receive a detailed error report within 48 hours, combining machine analysis with human expertise from architects and engineers.
It’s the kind of hybrid approach, equal parts industrial process and deep learning, that the founders hope could finally bring construction into the AI era.
The Pain Before the Build
For Peter Starr, Freeda’s American-born co-founder and CEO, the problem is more than theoretical.
“I spent three months in Abu Dhabi doing manual due diligence on plans during an internship,” he recalls. “It was mind-numbingly repetitive, checking for compliance line by line. And yet, the stakes were enormous. If a mistake slipped through, the cost was massive. I thought: there has to be a better way to do this.”
According to McKinsey, construction delays cost $1 trillion annually, and 70% of those delays start before construction even begins during the design phase.
Starr went on to graduate in urban planning at UCL in London, then studied business at HEC Paris, before joining AI sports media startup ScorePlay as Chief of Staff. There he met Augustin Perraud, now Freeda’s COO. “We realized how much potential AI had when applied to a real, physical-world problem,” says Perraud. “And in construction, the need is glaring.”
From Plans to Platform
The founding team is as international as the industry they’re trying to fix.
Alongside Starr and Perraud are Charles Desbaux (CTO, ex-Darktrace) and Mariano Rodriguez (CSO), a Cuban-born computer vision specialist who earned his PhD at France’s École Normale Supérieure.
Rodriguez’s research proved crucial. Architectural plans, he explained, are “layered documents – part geometry, part language, part regulation.” Traditional OCR or LLM models can’t process them properly.
Freeda’s solution merges computer vision and language models, interpreting not just symbols and text, but spatial logic – how rooms connect, how exits align, and how measurements translate to compliance.
“What we’re really doing is teaching the machine to think like a plan reviewer,” said Perraud. “It’s not about aesthetics. It’s about whether the building meets code, whether it’s accessible, and whether it’s safe.”
Freeda’s early traction spans the full construction ecosystem. Eiffage Construction, one of Europe’s largest builders, uses it to spot errors before they cause months of delay. Cohabs, a fast-scaling co-living operator, leans on Freeda’s expertise to coordinate plan reviews across international teams of architects and engineers. And an increasing number of retail and hospitality brands, ranging from quick-service restaurants to leisure chains, are adopting the platform to keep rollout schedules on track and ensure every site opens on time.
In 2025 alone, Freeda verified 10,000 plans, and aims to reach 1 million by 2026.
A New Kind of Construction Intelligence
Freeda’s model is refreshingly pragmatic. Instead of licensing the software, the company sells outcomes.
“Clients don’t pay for access; they pay for verified plans,” explained Starr. “If you’re a retailer opening new sites, or a real estate operator building residential projects, every month of delay is a month of lost revenue, sometimes €100,000 or more. We remove that risk.”
Framing Freeda from productivity tool to risk management platform may be what appealed to their investors most.
“Architectural plans are information designed for humans, but machines have always struggled with them,” said Pierre Entremont, partner at Frst, in the press release. “Freeda shows how AI can finally serve an industrial workflow, not replace it.”
For Brick & Mortar Ventures, a San Francisco-based construction tech specialist, Freeda embodies a new generation of applied AI. “This isn’t generic automation,” partner Guillaume Bazouin added. “It’s deep domain expertise embedded in a process that matters.”
Fast-Moving Foundations
If Freeda’s tech is deep, its momentum is even deeper. The startup was founded barely a year ago, in September 2024. By early 2025, it had its first working pilot. By June, the funding round was closed.
Since then, the team has grown from four founders to ten full-time employees, representing six nationalities. Most are engineers, from defense AI backgrounds to architecture specialists working out of Freeda’s Paris office at Station F, where the startup has been selected among the Future 40, the campus’s top 4% of startups.
Still, the team is careful not to scale sales faster than delivery. “We’re focused on making existing customers happy, growing those accounts, and improving accuracy before we flood the market,” said Perraud. “It’s about building credibility.”
The Ambition: Precision as Standard
The fresh capital will help Freeda hire more architects and field engineers to enrich its dataset, expand its AI engineering team, and adapt its platform to international regulations. The founders are clear on the challenge.
“Construction standards differ wildly from France to the US to the Middle East,” says Starr. “Our system needs to understand all of that, which means learning local practices, languages, and even drawing conventions.”
But the ambition is global. “What I say to the team is that within five years, I want Freeda to be used to build a 7-Eleven in Tokyo, a hospital in Riyadh, and a data center in Arizona,” Starr said with a smile. “Those are three completely different projects – small, medium, and huge – across three very different regions. That’s when we’ll know Freeda is really working.”
And if Freeda ever lands on the cover of The New York Times? Starr has that headline ready, too:
“We helped crack the housing crisis. If we can make building homes faster and more reliable by reducing errors and guaranteeing openings, then we’ve done something that truly matters.”
