Rowan Ramamurthy
Robotics engineer
Building the island
I build software that lets robots share work: multi-arm planners, drones and the tools that tie them to real machines.
This island is my timeline. Drive the AMR up the mountain, one stop per chapter. Or use the dots up top to jump ahead.
This island is my timeline. Scroll to drive the AMR up the mountain.
Off the clock I’m usually outside. Keep an eye out around the island for what I get up to; the AMR can even drive the seabed.
2017 to 2021
Control systems, autonomous navigation, computer vision, operating systems. Built an EKF SLAM, a UR5 pick-and-place cell with vision, and a C compiler along the way.
Plus a reef restoration team project: an engineered approach to coral seeding on the Great Barrier Reef, with a business plan to match.
2019 to 2020
Software for BLUEsat’s autonomous model Mars rover in the Off World Robotics competition. I integrated lidar SLAM for mapping and learned how ROS systems fit together across separate teams.
2019 to 2021
A top-mounted gimbal payload for the DARPA Subterranean Challenge: scanning and object detection in dark tunnels, independent of the drone’s own autonomy. Later, a 3-axis gimbal and light model in Gazebo and a port of the stack to the virtual challenge.
2021
An autonomous exploration framework that ranks features by how likely they are to be the point of the mission. Tested on partly explored telecom towers, it steers the drone to maximise point cloud density and completion where it counts.
2022 to 2024
Cirrus Real Time Processing Systems in Sydney builds the simulators Australian Navy and Air Force crews train on: tactical electronic warfare, the air combat officer trainer flown out of RAAF East Sale, and submarine and mine hunting trainers.
I wrote real-time simulation in C++ and C#. The highlight: an electromagnetic propagation model that predicts over-the-horizon coverage when signals duct through the atmosphere, with custom atmospheric profiles.
2024 to 2026
Master of Science in Robotics, Atlanta.
Deep reinforcement learning, multi-agent systems, manipulator mechanics and computer vision, plus a stint as a graduate teaching assistant.
2025
Hierarchical reinforcement learning for tendon-driven soft fingers in PyElastica. Each finger has its own low-level policy; a higher-level agent coordinates them to move an object.
2025 to 2026
My master’s research at Georgia Tech’s Advanced Manufacturing Pilot Facility: splitting one toolpath between several robot arms without them colliding.
The G-code is cut into contours with a layer dependency graph. Each free arm claims the best available contour; a cached ellipsoid prefilter with a reach check, and a learned pairwise model, skip most of the expensive collision checks. If an arm drops out mid-job, the schedule replans around it.
On the bench: two Tormach ZA6 arms on pedestals, drawing on a whiteboard level with their bases. The cell here runs the same idea live.
2026
I’m always up for talking robots, planners or a good line down a mountain.