I work on the autonomy stack behind industrial AGVs and AMRs — 3D LiDAR SLAM, multi-sensor calibration, and YOLO-based perception — plus the PLC and functional-safety integration that lets robots and people share a factory floor.
I'm a robotics engineer working on industrial autonomous vehicles — the AGVs and AMRs that move material through factories and warehouses. At VisionNav Robotics I work across the full autonomy stack: 3D LiDAR SLAM, multi-sensor calibration, perception, and the PLC and functional-safety integration that lets robots and people share the same floor.
My foundation is deliberately broad — an M.S. in Computer Science and a B.S. in Electrical Engineering, both from UIUC. That mix is the point: I'm as comfortable training and deploying a YOLO detection model (ONNX → TensorRT) as I am wiring Modbus TCP between an Allen-Bradley PLC and an AGV, or running FMEA on a heavy-load handling case.
Before robotics I shipped production software — full-stack web apps, React / React Native front-ends, and real-time data pipelines — which is why I think about robots as systems: perception, control, automation, and the upper-level scheduling and MES integration that ties them together. I'm drawn to problems where the success metric is physical: a robot that docks to the millimeter, a line that runs a full shift without intervention.
A working snapshot — grouped by where it sits in a robotics stack. Edit freely.
A video or GIF of the system running goes a long way — drop yours into each card's media slot.
My professional work bringing autonomous forklifts onto the factory floor — 3D LiDAR SLAM deployment, multi-LiDAR calibration, a YOLO vision pipeline, PLC integration, and functional-safety validation. A growing collection of what I've shipped here.
Four controllers designed in MATLAB/Simulink to balance an inverted reaction-wheel pendulum — from two- and three-state full-state feedback to a decoupled observer and an energy-based swing-up. The three-state design tolerated ~10% larger pulses and 50% larger disturbances than the rest.
A five-task manipulation routine on a robot arm — straight-line moves, 2-inch peg insertion, obstacle-avoiding zig-zags, and force-controlled egg pressing (500–1000 g) — built from task-space PD, inverse-dynamics, and impedance control with friction compensation, running at 1 kHz on a TMS320 DSP.
A 6-DOF arm that locates 3-block colored sticks from marker geometry (arcsin yaw), solves inverse kinematics with Joint-6 wraparound handling, and stacks two leaning towers into a stable 232 mm "Tower of Lire" bridge using overhang math and collision-aware waypoints.
A senior-design key dock that fuses ultra-wideband time-of-flight ranging (DWM1000) with an FSR pressure sensor and an ATtiny84A four-state machine. Includes a custom PCB, an LM1117 power chain, and an antenna-delay calibration analysis cutting ranging error from 30 cm to ~4.5 cm.
Hardware Flappy Bird on the Urbana FPGA board in SystemVerilog — a MicroBlaze + AXI system with a custom VGA controller (640×480), sprite ROMs, a START/GAME/INVINCIBLE/END FSM, and pixel-accurate multi-pillar collision. 8,089 LUTs, 55.5 MHz, 0.51 W.
Improved code reasoning and automated test generation on code LLMs (DeepSeek-Coder) through tailored prompt-engineering strategies.
Docker-based tooling to replicate research artifacts and standardize environments, making published research software reproducible.
Implementation of core cryptographic primitives — symmetric and asymmetric encryption, digital signatures, and blockchain fundamentals.
A React / Express web app that delivers personalized Steam game recommendations.
Statistical modeling to surface regional trends in global video-game sales.
I'm currently open to robotics and automation roles. If you have an interesting problem — or just want to talk shop — my inbox is open.