The job site stops lying when a computer is watching it.
Picture a half-built data center the size of three football fields. Hundreds of workers, dozens of trades, thousands of pipes and racks and conduit runs. Somebody, somewhere, is being asked the oldest question in construction: "Are we on schedule?" And somebody, somewhere, is about to guess.
Doxel exists so nobody has to guess. A worker walks the site with a 360-degree camera clipped to a hard hat. The footage uploads. By morning, Doxel's computer vision has measured what was actually installed - every visible trade, more than 80 stages of construction - and laid it against the schedule, the budget and the BIM model. Green means on track. Red means a problem that hasn't become a crisis yet.
Today Doxel runs on Fortune 500 job sites - healthcare campuses for Kaiser Permanente, energy projects for Shell, server farms for the developers racing to power the AI boom. It has measured more than three billion square feet of construction. That's who they are now. Getting here required betting that the most analog industry on earth was secretly a data problem.
Large construction projects are notorious for blowing past their deadlines and budgets. The villain is rarely a single dramatic failure. It's the slow accumulation of small ones - a trade that fell two days behind, an area that got billed for work that wasn't finished, rework nobody flagged until it was poured in concrete.
The catch: by the time these show up in a monthly progress report, they've already cost money. Status meetings ran on hand-built spreadsheets and optimistic estimates. Everyone reported percent-complete; almost nobody measured it. The truth lived in the field, and the field is loud, muddy and four weeks ahead of the paperwork.
Founder Saurabh Ladha saw this up close and took it personally. Delays, he points out, aren't just line items. A hospital that opens late is a hospital that isn't treating patients. He founded Doxel in 2015 after seeing how construction delays hit lives and livelihoods, not just balance sheets.
Ladha studied Management Science & Engineering at Stanford, where he got comfortable with cyber-physical systems - the unglamorous art of teaching software to understand the physical world. With co-founder Robin Singh, he made a contrarian wager: that you could quantify a construction site the way Google quantifies traffic, by reading reality directly instead of waiting for someone to report it.
The early version leaned on autonomous robots rolling through job sites with cameras and laser scanners - which is why Doxel's legal name still traces back to SR Autonomous, Inc. The robots were a clever answer. They were not, it turned out, the practical one. The team kept the computer vision and swapped the hardware for whatever was already on site: 360 cameras, drones, the occasional LiDAR scanner.
Stanford engineer who started Doxel after watching delays ripple into real lives. Forbes 30 Under 30; Goldman Sachs Top 100 Builders + Innovators.
Co-founded Doxel on the bet that construction's progress problem was, underneath everything, a computer vision problem waiting for the right model.
The pitch is easy to remember because it's accurate. Just as a maps app reroutes you around traffic before you hit it, Doxel flags cost and schedule risk before it becomes a delay. The harder part - the part competitors are still chasing - is the measurement underneath.
Walk the site with a 360 camera, fly a drone, or run a LiDAR scan. Doxel even added support for the rugged Insta360 X5, because job sites are hot, dusty and unkind to electronics.
Computer vision identifies every visible trade and 80+ stages of construction, then quantifies how much was installed, percent complete and the work rate.
Actual progress is matched against the schedule, budget and BIM model. Deviations surface as risk, not as a surprise in next month's meeting.
Teams validate pay applications, reconcile billing, and feed trade-level data into planning tools like Touchplan - so the schedule reflects reality, not hope.
Saurabh Ladha founds the company (as SR Autonomous) to measure construction with computer vision after seeing how delays hurt people, not just budgets.
Andreessen Horowitz and Amplo back the seed round, betting on AI for the analog construction world.
Doxel raises a Series A and leans into software-plus-camera capture over autonomous robots - cheaper, faster, everywhere.
Insight Partners leads a $40M round with a16z and Amplo, pushing total funding to $56.5M to scale the platform.
Partnerships with MOCA Systems (Touchplan) and Stream Data Centers position Doxel at the center of the AI-driven construction boom.
Skeptics are right to want evidence; construction is littered with software that demoed beautifully and died on a real job site. So here's the receipt that gets cited most: when Kaiser Permanente used Doxel on its Viewridge Medical Office project, the team delivered 38% more productively and came in 11% under budget.
The customer roster reads like a directory of people who cannot afford to be wrong about a deadline.
Doxel frames its mission around observability - giving construction teams a complete, honest view of their site, plus the tools to prevent delays, over-billing, re-work and trade stacking before any of them metastasize. The vision is bigger than dashboards: the company talks about the power of aligned teams to build the environment a thriving society depends on.
It's a tidy idea with a sharp edge. When everyone is looking at the same measured reality, the meeting gets shorter, the finger-pointing gets quieter, and the project gets built. Objective data has a way of ending arguments that opinions kept alive.
There's a neat irony in Doxel's 2025 trajectory. The same artificial intelligence that powers its computer vision is also driving a historic surge in data center construction - and those builds cannot slip. Every delayed megawatt is a delayed model, a delayed product, a delayed promise. Doxel now counts 14 of the top 20 data center developers among the teams it serves.
So the company that started by watching hospitals get built is now helping build the infrastructure of the AI economy itself. The tool that measures construction is being used to construct the machines that do the measuring. Recursive, slightly poetic, and very much on schedule.
Return to that half-built data center from the opening. Before Doxel, the answer to "are we on schedule?" was a confident shrug dressed up as a spreadsheet. Now it's a number - measured overnight, mapped against the plan, color-coded for whoever needs to act. The site stopped lying. That was always the whole point.