A hard hat. A 360 camera. An AI that knows what drywall looks like. Quietly, OpenSpace has turned the world's messiest industry into something you can scroll through.
It is 7:14 a.m. on a Tuesday and somewhere in Texas, a superintendent named Marco is clipping a 360 camera to his hat. He walks the third floor of a hospital under construction. He does not stop, he does not pose, he does not pin tags to anything. He just walks. By the time he gets to the elevator, the cloud already knows where every pipe, stud, and unfinished bathroom is. Eight time zones away, in a Mumbai design office, the project's MEP engineer is dragging a slider across the same floor - in November, then in March, then back to now. Nobody flew anywhere. Nobody had to.
This is what OpenSpace does, although the company would phrase it more politely. The pitch goes: spatial AI for construction. The reality goes: it took the oldest, dirtiest, most paper-bound industry on earth and gave it a save button. A walkable, time-stamped, AI-indexed digital twin of every jobsite, captured by literally walking around. Construction's first honest scoreboard.
The product is deceptively unglamorous. You buy an off-the-shelf 360 camera - the kind tourists use to film themselves on glaciers - mount it on a hard hat, and press record. OpenSpace's computer vision does the rest: stitching, pinning, comparing to BIM, flagging delta, generating something halfway between Google Street View and a court-admissible record. The company doesn't sell cameras. It sells the layer that makes the camera useful.
Construction is a $13 trillion industry that, for most of its history, has run on memory and Tuesday-morning standups. A wall goes up. Two weeks later somebody asks if the conduit was installed before the drywall. The answer is usually "I think so," followed by a phone call, followed by a flight, followed by tearing something out. Multiply that by every project on earth.
The founders of OpenSpace had spent enough time around construction sites to notice the size of the problem and around the Media Lab to notice the size of the opportunity. The world had cheap 360 cameras. The world had cloud computer vision. The world had drones. The world had BIM. The world did not have a way to plug all of it into one walkable record. Builders were sitting on petabytes of visual truth and using approximately none of it.
Here is the part that is mildly funny. Reality capture, as a category, sounds like the most boring thing a venture capitalist has ever funded. It is not robots. It is not chatbots. It is not whatever GenAI feature the latest deck is hyping. It is, in essence, photography with arithmetic. Which is exactly why nobody else saw it - and exactly why it works.
Jeevan Kalanithi, Michael Fleischman, and Philip DeCamp met at the MIT Media Lab, which is where you go if you'd like a graduate degree and a healthy disregard for what counts as a proper company. Kalanithi had already done his time - he co-founded Sifteo, which made interactive cubes that briefly lived in Apple stores, and sold it to 3D Robotics. He stayed at 3DR, ended up its president, and watched the company partner with Autodesk to put drones over construction sites. The numbers worked. The customers liked it. But the camera was always in the sky, and the building was always on the ground.
So in 2017, the three of them placed a bet that read, in retrospect, like a Mad Libs: take the cheapest commodity hardware available, attach it to the most low-tech possible operator, and use software to extract a category of data nobody knew was sitting on the table. Then sell it back to people who had been guessing about it for a hundred years. Lux Capital wrote the seed check. Menlo Ventures followed. Then Alkeon. Then a pile of strategics - JLL Spark, Navitas, Taronga, Nine Four - who knew the construction world and wanted in.
The bet, narrated charitably, was: software eats construction. Narrated honestly: someone is going to make the building industry legible, and we'd rather it be us than Procore.
OpenSpace doesn't ship one product anymore. It ships a stack. Each piece is sold as if it were a standalone tool, which is the polite SaaS way of doing things, but the real magic is what happens when they're stitched together.
The original. Walk the site with any 360 camera, drone, smartphone, or laser scanner. The platform auto-pins images to floor plans with no manual tagging.
Formerly ClearSight. The progress-tracking brain. Tracks 700+ visual components across 200+ schedule tasks and quantifies work-in-place from the captures.
Side-by-side BIM model and 360 image, with snap measurements. Catch coordination clashes before the trades catch them with a sledgehammer.
Drone capture for facades, earthwork, and the parts of a site that don't appreciate being walked.
Mobile photos, RFIs, punchlist items - auto-pinned to where they were taken. The death of "wait, which floor was that on?"
Most of the cleverness lives in what isn't a UI screenshot. The computer vision has to figure out where you are without GPS, because you are inside three feet of concrete. It has to recognize that the cluster of pipes in last Tuesday's capture is the same cluster, just now with insulation. It has to do this 10 billion square feet at a time. The CEO calls the underlying technology a "spatial AI engine," which sounds vague until you remember the alternative was a guy with a clipboard.
Construction software is supposed to take twenty years to scale. OpenSpace did the inconvenient thing and scaled in eight.
Kalanithi, Fleischman, DeCamp leave the orbit of 3D Robotics and the Media Lab.
$3.7M from Lux Capital. The 360-on-a-hard-hat hack becomes a product.
$14M led by Menlo Ventures. First general contractor customers go all-in.
$15.9M. Pandemic accelerates demand for remote jobsite oversight.
$55M from Alkeon. Platform expands beyond pure capture.
$102M led by PSP Growth at $902M valuation. Extended by $9M in August.
ClearSight rebrands as OpenSpace Track - the AI-tracking product gets a brain transplant.
The five products get repositioned as one system of work for builders.
OpenSpace's customer roster reads like a tour of who actually builds America: Suffolk, Level 10 Construction, JLL, Lee Kennedy, NOVO Construction. International accounts in Japan, Australia, the UK. The company is used on more than 14,000 jobsites and counting. The volume of imagery captured passed 10 billion square feet at last count, which is more than the floor area of Manhattan, repeatedly, by a wide margin.
The integrations matter too. OpenSpace plugs into Autodesk Construction Cloud and Procore - the two giants whose orbit any serious contech company has to live in. The cameras come from Insta360, Ricoh, GoPro. The drones from DJI and Skydio. OpenSpace is the connective tissue. Other people made the bones.
Ask Kalanithi what OpenSpace is for, and the answer drifts toward something almost philosophical. The built environment is the largest physical asset class on the planet. We live in it. We work in it. We pay an obscene amount of money to maintain it. And yet, for most of human history, we have known shockingly little about what's actually inside the walls until something goes wrong. OpenSpace's mission is to fix that - to make every square foot of the built world legible, queryable, and accountable.
Capture every wall Trust every photo Question every change orderThe vision compresses to a sentence: a construction industry where the digital twin is not an upsell, it's the baseline. Where insurance companies look at a captured record before they pay a claim. Where owners walk a building before it exists. Where the conduit was either installed before the drywall, or it wasn't, and there is a video to settle the argument.
It is fashionable, in 2026, to say that AI will eat every industry. Most industries are not actually that hungry. Construction is. The labor crunch is real. The schedule pressure is real. The capital cost of mistakes - rip-and-replace work, rework, lawsuits - is staggering. Every percentage point of efficiency translates to billions across the global market. OpenSpace sits exactly where the visual data flows in and the AI flows out. That's a useful piece of real estate.
What happens next is the interesting part. The captures pile up. The model gets smarter. The next product is probably less "look at the wall" and more "predict the wall." Which crews are productive on which scope. Which subs are about to slip. Which 8% of a project carries 80% of the rework risk. OpenSpace isn't there yet, but the data exhaust is starting to suggest where it's going.
Back to Marco in Texas. It is now 7:42 a.m. He has finished his walk. The camera is back on its tripod. The cloud has stitched, pinned, and indexed. By 8:00, the project manager in Boston has a daily report on her phone. By 8:15, the owner in Singapore has signed off on a milestone he's never seen in person. By 9:00, a coordination clash that would have cost $40,000 to demolish has been caught, flagged, and routed to the right sub. Marco didn't write a single email. He walked.
That is the part OpenSpace built. The walk was always free. The understanding was always the expensive bit.