Experience

Vecna Technologies

Robotics Engineer
Autonomy Stack

Robot Autonomy Stack

  • Integrated cameras, LiDAR, IMU & wheel encoders into a full localization + navigation stack.
  • Deployed AMCL-based localization for deployment in museums, aquariums & hospitals.
  • Implemented Extended Kalman Filter to fuse IMU and wheel odometry and a PID controller achieving 0.5 cm positional accuracy.
Go-to-Location Nav

Point to point navigation

  • Deployed a robust “Go-to-Location” feature with reliable localization, and dynamic replanning around unknown obstacles.
  • Designed real-time obstacle avoidance & dynamic replanning pipeline.
  • Achieved 94% goal-success across 60+ trials on three robots.
  • Demonstrated collision-free navigation in cafeteria, office lobby and museums.
Human-Following

Human-Following Robot

  • Integrated YOLOv11-based person detection with constant acceleration Kalman filter for robust human state estimation.
  • Implemented reactive control and recovery behaviors for smooth tracking under perception dropouts.
  • Validated reliable human-following performance in dense and highly dynamic crowds.
  • Achieved 89% tracking accuracy on straight, zig-zag & circular motion patterns.
  • Maintained mean following distance error of 2.2 ± 0.5 m.
  • Validated performance across multiple controlled trials.
Voice Commands

LLM-Based Voice Interface

  • Designed an LLM-driven voice interface for intent-aware robot control.
  • Mapped spoken commands to navigation goals and autonomy behaviors.
  • Enabled fully hands-free operation with wake-word–triggered commands including stop, go-to-location, follow-me, and docking.
Auto Docking

Fiducial Marker & AprilTag–Based Docking

  • Uses a pre-recorded AprilTag pose to localize and approach the docking station to a fixed offset.
  • Aligns the robot’s orientation relative to the tag before initiating a controlled docking maneuver.
  • Achieves reliable, hands-free autodocking via high-frequency (28 kHz) IR-based fine alignment.
System Design

System Architecture & Platform Migration

  • Led platform migration from NVIDIA Xavier to AGX Orin, including flashing, building, and validating standardized Docker images.
  • Ensured GPU compatibility and real-time performance for autonomy pipelines such as human-following and navigation.
  • Established a reproducible deployment workflow for cross-platform robotics development and testing.

NIOSH

Robotics Research Engineer
Memory-Augmented MPC

Predictive Control for Human-Aware Navigation

  • Developed a Memory-Augmented Model Predictive Controller (MAMPC) for human-following robots in dynamic, cluttered environments.
  • Incorporated short-term temporal memory to improve anticipation of human motion and intent.
  • Contributed to paper publication and experimental validation for an international robotics conference submission.
Masonry Robot Perception

Perception & Geometry for Unstructured Construction Tasks

  • Developed an AprilTag and fiducial-based perception pipeline, including plane orientation estimation relative to the ground.
  • Performed camera calibration and depth-based geometric reasoning using checkerboard patterns for reliable plane estimation.
  • Supported research on perception, control, and learning-based strategies for robotic masonry in unstructured environments.
  • Investigated gripper technologies (GelSight, GelSlim) for manipulation in construction contexts.
Human–Robot Proxemics

Safety & Proxemics Studies with Industrial Robots

  • Assisted in designing and executing human-subject experiments evaluating comfort and safety around industrial mobile robots (Freight100).
  • Configured multi-robot navigation tasks using ROS, Gazebo, AMCL, TF trees, and map servers.
  • Developed communication protocols to maintain fixed speed and separation distances during trials.
  • Supported motion-capture-based tracking of human and robot pose during experiments.
VR Human-Factors Lab

VR-Based Attention & Cognitive Load Experiments

  • Conducted VR-based experiments measuring human attention and cognitive load in simulated construction scenarios.
  • Collected and synchronized multimodal data from force plates, heart-rate sensors, and VR headsets across repeated trials.
  • Managed experimental flow, sensor calibration, participant onboarding, and reliable data logging.
  • Gained hands-on experience with ethical protocols and best practices for human-subject research.
Simulation & Testing

Simulation Environments & Experimental Pipelines

  • Set up Gazebo-based simulation environments with obstacles to test robotic masonry and navigation behaviors.
  • Developed 3D simulation test environments using gazebo and pybullet.
  • Collaborated on experimental design and data-collection procedures across multiple research projects.

Fluent Robotics Lab

Research Assistant — University of Michigan
Gaze-Based Intent Inference

Eye Gaze as a Predictor of Human Navigation

  • Analyzed MoGaze and THÖR-MAGNI datasets to study fixation patterns before locomotion decisions.
  • Modeled relationships between gaze direction, head orientation, and future motion intent.
  • Identified failure modes of gaze-only predictors in cluttered and ambiguous scenes.
  • Motivated latent goal inference beyond surface-level gaze trajectories.
Data Analysis

Dataset Processing

  • Built Python pipelines for synchronizing gaze, trajectory, and pose data.
  • Developed visualization tools for world-frame gaze alignment.

Flight dynamics Lab

Research Assistant - University of Michigan
Flight Testing

UAV Testing & Validation

  • Conducted flight testing on tailsitter and CRC custom UAV platforms.
  • Evaluated stability, control response, and mission execution reliability.
  • Iterated on tuning parameters using flight log analysis.
Mission Planning

Navigation & Position Control

  • Designed custom missions covering takeoff, waypoint navigation, and landing.
  • Implemented position-control strategies for precise waypoint tracking.
  • Validated navigation accuracy across repeated trials.
Autopilot Setup

PX4 Autopilot Configuration

  • Configured PX4 Autopilot on Ubuntu 22.04 using Docker-based deployment.
  • Built and validated firmware for custom UAV platforms.
  • Ensured reliable onboard system initialization and parameter consistency.
Ground Control

QGroundControl Integration

  • Integrated QGroundControl for real-time telemetry and mission monitoring.
  • Established stable communication via designated ports.
  • Debugged parameter synchronization and link reliability issues.
Flight Demo

PX4 Flight Demonstration

  • Tools: PX4 Autopilot, QGroundControl, ROS, Gazebo, C++.

ROAHM Lab

Research Assistant — University of Michigan
Reachability-Based Planning

Forward Reachable Sets

  • Worked on reachability-based trajectory planning using Forward Reachable Sets (FRS).
  • Integrated zonotope-based reachability analysis in C++, improving predictive trajectory computation for dynamic environments.
Planner Architecture

Modular Planner Design

  • Optimized the REFINE framework by implementing feedback linearization, enhancing trajectory planning and obstacle avoidance.
  • Reduced Forward Reachability Set (FRS) computation time by 58.6% through zonotope slicing, improving real-time predictive safety analysis.

Volvo Groups

Mechanical Engineer
Mechanical Design

CAD & Mechanism Design

  • Developed CAD models for AC gas cylinder transportation systems.
  • Designed tilting and handling mechanisms for safe material movement.
  • Validated designs for manufacturability and shop-floor deployment.
Automation Systems

AGV & 3MU Integration

  • Contributed to the design and deployment of Automated Guided Vehicle (AGV) systems.
  • Worked with 3MU automation units for material transport.
  • Supported system-level integration and testing.
Process Optimization

Operational Efficiency

  • Applied Spaghetti diagram analysis to identify workflow inefficiencies.
  • Redesigned shop-floor layouts to reduce unnecessary motion.
  • Improved throughput and material handling efficiency.
Digital Andon

Shop-Floor Monitoring

  • Led deployment of a digital Andon system for real-time status tracking.
  • Enabled faster issue detection and response on the shop floor.
  • Improved visibility into production bottlenecks.
Project Visuals

Industrial Logistics Setup

Efficiency Enhancement in Industrial Logistics
  • Tools: Creo, CAD, 3D Automation Software.