Kushant Patel spent two years at the University of California, Berkeley teaching quadruped robots to map and inspect nuclear material. He now develops control systems for advanced synchrotron beamlines at Brookhaven National Laboratory. On 9 October 2026 he discussed robots in scientific facilities, such as particle accelerators, in Episode 165 of the Robot Talk podcast, in conversation with the host, Claire.
A career spent on perception and motion
Patel describes his work as building software that helps robots and instruments perceive, map and move through the physical world. That description covers each stage of his career so far, and it sets the frame for how he approaches machines that operate in specialised research settings.
He completed a Master's degree in Mechatronics Engineering at the University of Waterloo. After that he moved to the University of California, Berkeley, where he spent two years working with quadruped robots. The task was to teach them to map and inspect nuclear material. That is a job where the environment is hazardous and the robot has to build an accurate picture of its surroundings before it can do anything useful in them.
His next post was at the University of California, Davis, where he spent a year building closed-loop brain-computer interfaces. A closed-loop system is one where the output is continuously fed back into the input, so the software has to sense, decide and act as a single cycle. The same pattern runs through robot control, which is one reason the move from neural interfaces to legged robots and then to instrument control is less abrupt than it might look.
The work at Brookhaven
At Brookhaven National Laboratory, Patel is developing control systems for advanced synchrotron beamlines. A beamline is the part of a synchrotron facility where a beam is directed at samples and measured. The source describes his role as control systems for these beamlines, and it does not give details of the specific hardware, the number of robots involved or the performance targets. Those details are not stated, so they are not reported here.
What can be said is how the work fits his stated focus. Instruments in a facility like this must be positioned, aligned and monitored, and the same perception, mapping and motion problems he worked on at Berkeley apply when the machine is a robot arm or an instrument stage rather than a walking robot. The episode frames the subject as robots that work in scientific facilities, with particle accelerators given as an example of that kind of site.
What the conversation covered
The episode is a conversation between Claire and Patel about the robots behind scientific research. The summary of the episode names the central topic as robots that work in scientific facilities. It does not list individual questions, quotes or findings, so this report does not attribute any specific claims to Patel beyond his background and his current role.
Robot Talk is a weekly podcast about robotics, artificial intelligence and autonomous machines. Episode 165 sits in a run of episodes that includes conversations on construction robotics and on evolving robot bodies and brains, which shows the series covers a wide range of applications rather than a single sector.
Why this matters for the field
The analysis in this section is our own reasoning from the facts above, not a claim made in the episode.
Most public attention on robotics goes to humanoids, warehouse machines and consumer products. Robots in scientific facilities get less coverage, yet Patel's career shows how much of the underlying technology is shared. Mapping and inspection with quadrupeds, closed-loop control in brain-computer interfaces and beamline control all depend on the same loop: sense the world, build a model, act, and check the result. A researcher who has worked across all three is evidence that these skills transfer between fields that look unrelated from the outside.
There is also a practical point about where the demand sits. Facilities such as synchrotrons and accelerators are environments where access can be restricted and where the equipment is specialised. His Berkeley work involved nuclear material, which is a setting where sending a robot in place of a person is an obvious benefit. His Brookhaven work involves control systems for beamlines, which suggests that in research infrastructure, software that coordinates instruments is as important as the robots themselves. That is an inference from his job description, not a finding reported in the episode.
Comparison with other robotics sectors is limited by what the source provides. Industrial and logistics robots usually have a clear commercial metric, such as throughput. The source gives no equivalent measure for scientific facilities, such as experiment time saved, samples handled or downtime reduced. Without such figures, it is not possible to say how far robot-driven control has progressed at Brookhaven or how it compares with other laboratories.
Several things remain unproven or unstated. The source does not say whether the beamline control systems Patel is developing are deployed, in testing or still at the design stage. It does not name the robots used, report any results from the Berkeley inspection work, or say how the quadrupeds performed. It does not describe what, if anything, from the brain-computer interface work carries over to his current projects. Anyone assessing the state of robotics in research facilities should treat the episode as a profile of a practitioner and a pointer to the topic, not as evidence of measured progress.
What the episode does establish is that robotics in science is being built by engineers with deliberately broad experience, and that the field is treating facilities like particle accelerators as a robotics problem in their own right. Whether that translates into published benchmarks or shared tools is the open question to watch.