Bedrock Robotics

Boris Sofman joined Xora Innovation's Origin Stories to talk about the difference between building technology in a lab and deploying it in the field, and what it takes to make autonomous systems reliable when conditions are unpredictable and the stakes are real. Watch the full interview below.

Construction workers build the infrastructure that everything else depends on. They deserve careers that don't cost them their health. Omar Percy, CHST from Sundt Construction shares with Linda Xu why he believes autonomous technology could mean less fatigue, less pressure, and better work-life balance for the crews doing this work.

The future of autonomous technology in construction is no longer far off. It is here, and it is moving fast.
 Today’s conversation with the Bedrock Robotics team reinforced how quickly autonomy, robotics, and data-driven tools are beginning to reshape the way we build.
 For our industry, the opportunity is clear: Safer jobsites, greater precision, more predictable schedules, and better visibility into how work is progressing in the field.
 At Clayco, we are focused on investing in the technologies and partnerships that help our team build smarter, safer, and faster for our clients.

From a children's birthday party in a highway median, to a semi-truck struck by lightning, one of our team's biggest learnings from putting driverless cars on freeways is that you never know what's going to surprise you next. Robotics in the real world is like that. The environment doesn't cooperate, and the surprises don't stop coming. The job is to encounter them safely, address them, and build a system that handles whatever comes next. Construction is no different. And that's exactly why we put machines on the ground early, because the only way to find the surprises you haven't imagined yet is to show up and meet them. Field time is how we build the library of edge cases that makes the system smarter over time. If that kind of problem sounds like the one you want to work on, we're hiring. Link in the comments.

"Why would we not want some sort of a digital operator in construction equipment?" Omar Percy, CHST, Group HS&E Manager at Sundt Construction, wasn't always a believer. But watching autonomous vehicles operate safely on public roads shifted something. The logic started to hold: if a digital operator can deliver consistent, repeatable outcomes in one high-stakes environment, it can probably do it in another.

A case study went live today on how we built Bedrock Robotics ML infrastructure with Anyscale & ray - a lot of this is Thomas Pelletier's work with the Bedrock Robotics infrastructure team. Our pipelines are diverse - a single job might decode video on GPUs, run analysis on CPUs, then run model inference and simulation on GPUs again. Managing that with a focused platform team supporting dozens of researchers and all our engineers across the company is a meaningful constraint – and a challenge we knew would help us produce great systems.   We also believe that efficient, observable, and well structured infrastructure is important when we’re early in our journey and is a superpower as we grow, and we are committed to learning by solving those hard problems. We used Anyscale & ray to help us leapfrog us to efficiency and functionality - touching on everything from maximizing spot use (70%-80%) to workload scheduling to dramatically growing our training pipelines (millions of hours/month) over the first 2 years of the company. If you want to read a bit more about what we did, we published a case study with Anyscale. https://lnkd.in/ePvBimt3

Robotics is having a category-defining moment, and the communications work happening inside these companies is shaping how an entire industry is understood. This Thursday in New York, our Head of Marketing and Communications Anjelica Triola Sarkar joins Alyssa Demirjian (Viam) and Katheryn Thayer (Radical AI) for a conversation on what it takes to do this work well, from managing public perception to building teams that can tell highly technical stories clearly.

This is what the Bedrock Operator sees on a job site. The camera feed shows what a human operator would see. The LiDAR point cloud shows what the machine knows, the full geometry of the environment, rendered in real time, down to millimeter precision. The excavator's position, the bucket's angle, the shape of the cut: all of it mapped and updated continuously. This is the sensing foundation that makes autonomous operation possible. Before a machine can act safely and accurately, it has to understand its environment completely. 

Before an autonomous machine can act, it has to understand. A construction site is one of the most unpredictable environments imaginable. The terrain changes with every pass. Material piles shift. Other machines move nearby. No two digs are the same. For a human operator, reading that environment is second nature, built from years of experience and constant sensory feedback. Teaching a machine to do the same is one of the hardest problems in autonomy. Scene understanding is how we're working to solve it. Our system fuses LiDAR point clouds, camera feeds, and real-time terrain data into a continuous, three-dimensional picture of everything happening around the machine. It's a living model of the environment, updated moment to moment so the machine can make decisions grounded in what's actually in front of it.