Path Robotics turns blind pull requests into pre-validated code with RWX
Path switched from a slow, resource-heavy CI pipeline to RWX, cutting feedback loop time by 83% while eliminating redundant downloads.
"The most important job of a CI pipeline is to give developers timely feedback on code quality and whether something is ready to ship. RWX has optimized this and made a platform that delivers feedback much sooner than most other platforms even provide any feedback."
About
Path Robotics is a technology company that offers advanced robotic welding systems to manufacturing and industrial businesses. Leveraging artificial intelligence, its robots are designed to adapt and self-program to perform precise welding operations. Backed by $380M in funding, Path strives to make sophisticated automation technology accessible to businesses of various sizes.
Challenge
Uncached dependencies and long queues blocked fast developer feedback
Path builds autonomous welding robots for carbon steel manufacturing, helping industrial businesses drive efficiency while addressing persistent, industry-wide labor shortages. Naturally, shipping reliable, high-quality code is core to keeping each fleet of robots running in the field. But as Path's platform grew, its CI pipeline shifted from a quality gate to a blocker.
At the time, Path's software ran in Docker containers and depended on heavy robotics libraries and computer-vision packages to build. With no caching between builds, every single PR re-downloaded the same dependencies from scratch, burning 20-30 minutes before a single line of new code was even evaluated. For Josiah Lansford and Garrison Bybee, platform engineers at Path who oversee cloud infrastructure and software deployment, having urgent fixes stuck behind the same delay as routine work was not sustainable.
Flakiness was another persistent problem. Test failures weren't always caused by committed code. Bad builds or tests could fail on their own. As a result, developers had to re-run jobs and investigate whether a failure was real or a fluke. The feedback loop consequently became even more time-consuming, with PRs spending hours in the queue before getting merged. "Our software can be very flaky when it comes to building and testing," Garrison explains. "Builds could take an hour, and tests an extra 45 mins, and we would do this twice."
For Josiah and Garrison, the priority was clear: find a CI/CD solution compatible with their existing stack while delivering faster and more reliable feedback to developers. After discovering RWX, they were confident it could reshape how their team shipped software, without paying the same dependency tax on every PR.
"The feedback loop from a developer's change to actual review was extremely long. You'd be approved and sitting in the queue, but still waiting four hours to merge, so if we needed to get changes in to cut a release, there was a lot of time spent just waiting around."
Solution
RWX replaces full rebuilds with cached, reusable outputs
Rather than handing Josiah and Garrison an off-the-shelf solution, RWX took ownership of integrating the platform within Path's existing stack. When Path's images required a specific file path that didn't match RWX's default setup, the RWX team built a custom entry point to support it and shipped the fix within a week of Garrison flagging it. "RWX was very receptive to our unique needs and shipped new customization features that made the migration even easier than it had to be," Josiah shares.
Migrating an existing repository onto RWX took Josiah less than an hour, thanks to documentation clear enough to follow without prior context. Now, RWX caches Path's dependency tree instead of rebuilding it from scratch on every pull request, ensuring a build only pays for what's actually changed. Using RWX's CLI, engineers can kick off and validate a full pipeline run from their own machine before ever opening a PR, catching problems in real time instead of waiting on a queued CI run.
Builds that once took up to two hours, largely spent re-downloading the same robotics and computer-vision dependencies on every single PR, now typically finish in about 20 minutes. Because validation happens before code reaches review, engineers know their work is ready to ship well before a reviewer ever sees it, replacing a workflow where failures often surfaced only after a PR was already open.
Path's code validation process has also become easier to trust. "RWX just provides a cleaner way to see where the failure is," Garrison says. Combined with improved platform stability and continued support from the RWX team, that visibility has made it easier for developers to act on CI results with confidence.
"RWX turned our hour build time into 20 minutes, max. Sometimes it finishes in 10 minutes, and sometimes five. That's what makes it so good."
Results
From two-hour builds to a 20-minute feedback loop with RWX
With RWX, Path unlocked a caching layer built to absorb its heaviest dependencies without re-downloading them on every single PR. Engineers now get feedback in minutes instead of hours, while Josiah and Garrison can rely on the platform to handle build and test load without babysitting a fragile pipeline.
Looking ahead, Path is excited about RWX's AI testing workflow, which lets agents write and validate code independently through the CLI. As the bottleneck in software development shifts from writing code to validating it, its team sees RWX as a solution built exactly for that moment.
"One feature I'm particularly excited about is RWX's AI testing workflow. For Path, the bottleneck moved from writing to validating code, and this feature is designed to make that validation more effective and scalable."
Never miss an update
Get the latest releases and news about RWX with our newsletter.
More customer stories
“The CLI and other features built into RWX continue to be some of the best parts of the platform. It is designed in a way that really enables fast feedback loops with agentic help.”
“RWX didn't just make our pipelines faster; it changed how we think about delivery. We can ship whenever we want, trust our tests, and start layering AI on top of a platform that already runs our entire stack.”
“The past week testing out our CI runs in RWX has been beautiful. The speed, the ability to see failures coming in in realtime, and then being able to retry just the failed specs.”
“RWX just worked right from the beginning, and still does today. Even through GitHub outages, we haven’t had to deal with it at all, which is amazing.”
“RWX is super fast. It executes jobs quickly, caching works incredibly well, and it’s all very hands-off. But most importantly, the cost savings have held as our usage has scaled.”
“Our CI was slow, and with that long execution time came some noticeable shifts in developer willingness to innovate. RWX’s ability to reuse prior results when the underlying inputs had not changed gave us a way to break that trend.”
“We needed to reliably get our code shipped to production quickly, every time. RWX made that possible by giving our engineers a feedback loop they could actually rely on.”
“The CLI and other features built into RWX continue to be some of the best parts of the platform. It is designed in a way that really enables fast feedback loops with agentic help.”
“RWX didn't just make our pipelines faster; it changed how we think about delivery. We can ship whenever we want, trust our tests, and start layering AI on top of a platform that already runs our entire stack.”
“The past week testing out our CI runs in RWX has been beautiful. The speed, the ability to see failures coming in in realtime, and then being able to retry just the failed specs.”
“RWX just worked right from the beginning, and still does today. Even through GitHub outages, we haven’t had to deal with it at all, which is amazing.”
“RWX is super fast. It executes jobs quickly, caching works incredibly well, and it’s all very hands-off. But most importantly, the cost savings have held as our usage has scaled.”
“Our CI was slow, and with that long execution time came some noticeable shifts in developer willingness to innovate. RWX’s ability to reuse prior results when the underlying inputs had not changed gave us a way to break that trend.”