Kulunu Samarawickrama

I am Kulunu Samarawickrama, a Computer Vision and Robotics AI specialist based in Finland. I build software and research workflows that help robotic and intelligent systems understand visual and 3D sensor data. My work connects computer vision, machine learning, robotics software, and experimental validation, with a focus on making perception systems reliable enough to test, explain, and improve.

I have more than five years of experience working with real and simulated sensor data for robotic perception, manipulation, and intelligent system evaluation. My background includes 3D computer vision, RGB-D perception, point-cloud processing, segmentation, object detection, and 6-DoF pose estimation. I am especially interested in systems where cameras, depth sensors, robot models, datasets, and evaluation tools need to work together as one practical workflow.

I work mainly with Python, C++, PyTorch, PyTorch Lightning, ROS 2, Gazebo/Ignition, OpenCV, Open3D, PyTorch3D, Linux, Git, Docker, Apptainer, and SLURM/HPC environments. I have built synthetic and real-data pipelines for robotic perception, including CAD and simulation-based data generation, controlled dataset splits, benchmarking, and error analysis.

My work is not limited to model training. I also build the surrounding tools that make AI systems useful: data preparation scripts, simulation setups, sensor-processing pipelines, evaluation reports, documentation, and reproducible experiments. This helps connect perception models with robot behaviour, sensor quality, and real engineering constraints.

Alongside my technical work, I bring published research experience, open-source project work, technical writing, project documentation, progress reporting, publication preparation, and MSc thesis supervision. I am comfortable working in multidisciplinary teams where AI models, robotic systems, data quality, and experimental reliability all need to connect clearly.

Expertise

• Computer vision: detection, segmentation, RGB-D perception, point clouds, and 6-DoF pose estimation
• Robotics AI: perception pipelines, manipulation workflows, robot sensing, and simulation-based evaluation
• Machine learning: PyTorch-based model development, training workflows, benchmarking, and error analysis
• Robotics software: ROS 2, Gazebo/Ignition, sensor integration, CAD/URDF models, and Linux-based tools
• Data workflows: synthetic and real-data pipelines, dataset design, controlled splits, and reproducible experiments
• Technical communication: documentation, research writing, progress reviews, publication preparation, and supervision

I am open to collaboration across computer vision, robotics AI, applied machine learning, robotic perception, research engineering, and intelligent systems, especially where real-world sensor data and reliable AI workflows are central.