Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Hi, I’m Mehrnoosh Afshar, a robotics and ML engineer focused on end-to-end perception-to-control systems for real robots. I design and deploy perception, planning, and control pipelines, with hands-on experience in motion planning, video segmentation/tracking, and closed-loop controllers. I work with world models and MPC/MBRL for process control, and I’m proficient in Python, C++, ROS 2, and PyTorch with TensorRT integration. In industry and academia, I’ve built production-ready stacks, mentored co-op students, and contributed to sim-to-real tissue robotics and welding analytics. I’ve deployed AI-driven perception to production welding robots, improved frame segmentation with foundation-model distillation, and developed real-time control loops and model-based planning. I enjoy collaborating across disciplines to solve challenging robotics problems and bringing research ideas into real-world impact.…Hi, I’m Mehrnoosh Afshar, a robotics and ML engineer focused on end-to-end perception-to-control systems for real robots. I design and deploy perception, planning, and control pipelines, with hands-on experience in motion planning, video segmentation/tracking, and closed-loop controllers. I work with world models and MPC/MBRL for process control, and I’m proficient in Python, C++, ROS 2, and PyTorch with TensorRT integration. In industry and academia, I’ve built production-ready stacks, mentored co-op students, and contributed to sim-to-real tissue robotics and welding analytics. I’ve deployed AI-driven perception to production welding robots, improved frame segmentation with foundation-model distillation, and developed real-time control loops and model-based planning. I enjoy collaborating across disciplines to solve challenging robotics problems and bringing research ideas into real-world impact.

Mehrnoosh Afshar

Data Scientist, AI Engineer, AI Developer, +1





Hi, I’m Mehrnoosh Afshar, a robotics and ML engineer focused on end-to-end perception-to-control systems for real robots. I design and deploy perception, planning, and control pipelines, with hands-on experience in motion planning, video segmentation/tracking, and closed-loop controllers. I work with world models and MPC/MBRL for process control, and I’m proficient in Python, C++, ROS 2, and PyTorch with TensorRT integration. In industry and academia, I’ve built production-ready stacks, mentored co-op students, and contributed to sim-to-real tissue robotics and welding analytics. I’ve deployed AI-driven perception to production welding robots, improved frame segmentation with foundation-model distillation, and developed real-time control loops and model-based planning. I enjoy collaborating across disciplines to solve challenging robotics problems and bringing research ideas into real-world impact.…Hi, I’m Mehrnoosh Afshar, a robotics and ML engineer focused on end-to-end perception-to-control systems for real robots. I design and deploy perception, planning, and control pipelines, with hands-on experience in motion planning, video segmentation/tracking, and closed-loop controllers. I work with world models and MPC/MBRL for process control, and I’m proficient in Python, C++, ROS 2, and PyTorch with TensorRT integration. In industry and academia, I’ve built production-ready stacks, mentored co-op students, and contributed to sim-to-real tissue robotics and welding analytics. I’ve deployed AI-driven perception to production welding robots, improved frame segmentation with foundation-model distillation, and developed real-time control loops and model-based planning. I enjoy collaborating across disciplines to solve challenging robotics problems and bringing research ideas into real-world impact.

Available to hire

Hi, I’m Mehrnoosh Afshar, a robotics and ML engineer focused on end-to-end perception-to-control systems for real robots. I design and deploy perception, planning, and control pipelines, with hands-on experience in motion planning, video segmentation/tracking, and closed-loop controllers. I work with world models and MPC/MBRL for process control, and I’m proficient in Python, C++, ROS 2, and PyTorch with TensorRT integration.

In industry and academia, I’ve built production-ready stacks, mentored co-op students, and contributed to sim-to-real tissue robotics and welding analytics. I’ve deployed AI-driven perception to production welding robots, improved frame segmentation with foundation-model distillation, and developed real-time control loops and model-based planning. I enjoy collaborating across disciplines to solve challenging robotics problems and bringing research ideas into real-world impact.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Language

English

Fluent

Work Experience

Machine Learning Engineer & Roboticist at Novarc Technologies

June 1, 2023 - Present

Designed and deployed motion-planning solutions with MoveIt 2 and cuMotion; developed an industrial planner within the MoveIt 2 framework, validated in NVIDIA Isaac Sim, and successfully transferred to production welding robots. Implemented point-cloud processing and part localization to align CAD-based planning scenes with real-world robot cells in simulation and on hardware. Built a vision-based perception stack for welding, enabling arm guidance and online regulation of welding parameters; architected two neural components for frame segmentation and mask tracking. Developed multiple segmentation models using ViT and distilled knowledge from foundation models (e.g., Segment Anything, DINO) to improve accuracy and robustness. Advanced video segmentation to enhance frame-to-frame consistency and reduce false positives under disturbances (e.g., spatter, glare). Fine-tuned diffusion-based image generation models, conditioning on masks to synthesize high-quality training data and boost se

Associate Machine Learning Developer at AltaML

January 1, 2023 - April 1, 2023

Built a pipeline to predict unexpected equipment downtime one hour in advance from sensor data and identified failure causes to provide actionable insights.

Research Assistant at University of Alberta

January 1, 2019 - September 1, 2023

Developed a sim-to-real tissue simulator by combining a graph neural autoencoder (GNN-VAE) with data assimilation; used SOFA for tissue modeling and Python/PyTorch for the learning and sim-to-real modules. Created a graph-based autoencoder to extract deformation components of mesh-represented objects. Built a mechatronic system including a real-time tissue deformation solver and a model-based MPC controller for deformation control during breast brachytherapy; implemented in C++ and MATLAB. Developed a hybrid tissue deformation solver by combining a real-time solver with a Kalman filter to fuse optical measurements and adaptively correct model error. Implemented autonomous needle-tip tracking using a Panda robotic arm with a vision-based marker tracker; developed in MATLAB/Simulink and ROS. Built a system to control Panda robots via hand gestures by recognizing movements in real-time video streams. Designed and constructed a new Remote Center of Motion (RCM) mechanism for ultrasound pro