Soroush M.

1- Developed a ROS2-based navigation solution for outdoor autonomous navigation for golf ball picking. 2- Integrated and configured Nav2 to enable path planning, controls, and behavior policies of the robot using custom behavior trees, achieving an average lateral distance error of 12 cm over several-kilometer-long navigation missions. 3- Created modular Docker images to deploy software onboard. 4- Improved the path tracking accuracy of the controls algorithm by introducing curvature-aware cost terms for sloped, slippery path following of the mobile robot. 5-Architected and developed a maintainable sensory healthcheck module for the pikr robot, capable of detecting and reporting several sensory faults and failures per sensing system for reliable runtime operation of the robot. 6- Designed overall software architecture and led code reviews. 7- Helped envision Gazebo simulation environment for the ball picking robot.

1- Optimized LiDAR‑inertial SLAM accuracy by integrating LIO‑SAM, eliminating drift in subterranean autonomous mapping. 2- Developed real‑time IMU‑LiDAR sensor fusion algorithms for robotic localization and mapping, achieving positional accuracy within <3 cm. 3- Implemented DevOps workflows in GitHub Actions for automated builds, testing, and deployment, reducing CI/CD pipeline runtime by 40%

1- Accelerated a CNN‑based semantic segmentation network using TensorRT in C++ API, boosting inference speed by a factor of 20. 2- Deployed multiple computer vision models including pose estimation, instance segmentation and object detection, to Nvidia Jetson Orin. 3- Developed a multi‑sensor vision pipeline for seamless time of flight camera sensor integration and synchronization in C++ using an OOP paradigm. 4- Engineered and implemented a high-performance multi-threaded AI handling server in C++, facilitating the integration of sensor data.

Worked in an agile team of data scientists to: 1- Carry out statistical distribution analysis and predictive model development of internet speed data in a comparative validation analysis based on Ookla and StatCan datasets. 2- Performed dataset cleaning and merging as well as ML-based model development to estimate material properties, based on 30 years of recorded data from Metro Technical Services co. to improve material property estimation accuracy. 3- Implemented vision-based anomaly inspection algorithms for concrete, decreasing inspection times by %90 in construction. 4- Deployed a time-series forecasting model for the task of manpower prediction of construction workers with an accuracy of ± 5 people/week.

1- Developed vision‑based 3D object localization algorithms for novel state estimation in Python, using OpenCV, Pytorch, and NumPy. 2- Integrated the developed algorithms for the task of Automatic Material Deposition by testing on real robots at AHTST lab. 3- Secured more than $3000 worth of discount by objectively negotiating with the MOTOMAN company, to receive a software development kit. 4- Supervised and led 2 undergraduate students and 1 high school intern at the NODE lab by defining projects, deliverables, and timelines. 5- Employed the YOLO, OrienMask, and CoEx frameworks for the task of stereo 3D object detection and localization. 6- Implemented the Python code to deploy 3D point cloud manipulation, statistical filters, and Kalman optimal variance filters for state estimation. 7- Incorporated onboard and off‑board cameras to refine 3D object pose estimation through designed cost functions via optimization techniques. 8- Developed real-time MPC for the task of custom trajectory tracking for an industrial heavy duty mobile robot.

Artificial Intelligence Statics of Solid Bodies Engineering Statistics Material Sciences

Participated in the autonomous driving systems design team and developed camera calibration code, tested successfully in CARLA simulator. • Acquired proficiency in deep convolutional neural networks, sensor fusion, and state estimation techniques, commonly used in autonomous navigation and self‑driving cars.

Conceptualized and modeled a rover robot that was to compete in the canceled agricultural robot competition. • Designed and built a light box for quality assurance of batteries using SOLIDWORKS as a side‑task to generate income for the team. • Performed extensive experiments to improve the design parameters of the SSL robot solenoids for the ball‑throwing task using Arduino.

Research Assistant at NODE Lab. Computer Vision, Novel State Estimation, Automatic Navigation, Mobile robots, Controls Engineering, Artificial Intelligence, Machine Learning