Adaptive Gait Controller
Model-predictive controller for quadrupedal robots that adapts stride frequency and foot placement in real time based on terrain estimation from onboard IMU and depth data.
// robotics engineer & researcher
I build systems that turn perception into motion — legged robots, manipulation pipelines, and the control stacks that make them move like they mean it.
I'm a robotics engineer focused on the intersection of control theory, perception, and mechanical design. My work spans legged locomotion, dexterous manipulation, and the software stacks that let hardware behave predictably in unpredictable environments.
Currently based in San Francisco, CA, working on real-time control systems for autonomous platforms. Previously built manipulation pipelines for warehouse robotics and contributed to open-source motion planning tools.
I care about robots that fail gracefully, code that reads like documentation, and systems where the math and the mechanism agree with each other.
Leading control systems development for a quadrupedal research platform. Designed the real-time state estimation pipeline and rewrote the gait scheduler, cutting recovery time after disturbance by 40%.
Built perception-to-grasp pipelines for warehouse pick-and-place systems. Integrated depth cameras with a learned grasp-planning model, deployed across 30+ production cells.
Researched sim-to-real transfer for legged locomotion policies. Published findings on domain randomization strategies that reduced the sim-to-real gap on rough terrain.
Focused coursework on dynamics, controls, and mechatronics. Team lead for the university's autonomous rover competition entry, placing top 5 nationally.
Model-predictive controller for quadrupedal robots that adapts stride frequency and foot placement in real time based on terrain estimation from onboard IMU and depth data.
End-to-end pipeline combining point-cloud segmentation with a learned grasp-quality network, deployed on a 6-DOF arm for unstructured bin-picking tasks.
Custom VIO stack tuned for compute-constrained embedded platforms, running at 60Hz on a Jetson Orin Nano with sub-2% drift over 500m indoor traverses.
Open-source benchmark for evaluating domain-randomization strategies across five legged-robot morphologies, with reproducible baselines and terrain generators.
A. Rivera, et al. — IEEE International Conference on Robotics and Automation (ICRA)
A. Rivera, et al. — IEEE/RSJ International Conference on Intelligent Robots (IROS)
A. Rivera, et al. — Robotics: Science and Systems (RSS) Workshop