Construction has mechanised much of the heavy work outside buildings, but a surprising amount of interior finishing is still done manually. Origin wants to change that equation with autonomous robots — and has emerged from 18 months of stealth with $20 million in funding to scale the idea.
The round was co-led by Accel and Zacua Ventures, with participation from Unilazer Ventures, Schema Ventures and Z21, alongside angel investors associated with companies including Anduril, Skild AI, 1X and Tesla. The capital will primarily go toward expanding Origin’s fleet of OG-1 construction robots and running customer pilots across international markets.
Founded in early 2025 by Yogesh Ghaturle, Ashish Daga and Tushar Makkar, Origin brings together startup, robotics and AI experience. Ghaturle previously founded Bengaluru retail-tech startup Perpule, which Amazon acquired in 2021. Makkar leads autonomy at Origin, while Daga heads hardware development.
Their first product, OG-1, is built for interior wall finishing. The mobile robot can apply compound, sand and prime walls, work at heights of up to 14 feet and operate alongside construction crews rather than inside a segregated robotic zone. Origin says the system is already performing wall-finishing work at roughly three times human speed while maintaining human-level quality.
Behind OG-1 is what Origin calls its Construction Action Model, which handles task planning, movement and feedback-based corrections on changing job sites. The company has already deployed robots across offices, multifamily housing, data centres, hotels and warehouses, and plans to expand the same platform into areas such as taping, drywall installation, drilling and conduit work.
Origin is targeting a fleet of 150 deployed robots by the end of 2027, with pilots planned across the US and Europe. The larger bet is that physical AI can finally automate the messy, variable interior construction work that traditional industrial robots were never built to handle.
Filed by
Startup Unplugged

%20(1).png)
%20(1).png)