I took a career break to fully focus on studying deep learning and computer vision in order to fill knowledge gaps and transition to a Deep Learning (Computer Vision) Engineer role. Recently, I've also started writing about my learnings and personal projects at https://abojda.github.io𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 𝐬𝐭𝐚𝐜𝐤:- Python- PyTorch, PyTorch Lightning, TIMM, torchvision- PIL, OpenCV, optuna, albumentations, Weights & Biases, - numpy, pandas, scikit-learn, matplotlib, seaborn- Became proficient with deep learning frameworks, including PyTorch and PyTorch Lightning- Developed and deployed state-of-the-art deep learning models, with a focus on solving computer visionproblems, utilizing a diverse range of training techniques, including transfer learning and self-supervisedlearning- Implemented effective data augmentation strategies and conducted comprehensive hyperparameter tuning to optimize model performance and enhance generalization capabilities- Acquired extensive knowledge in both classical machine learning (supervised and unsupervised) and advanced deep learning techniques, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformers, and Generative Adversarial Networks (GANs)

I took part in two main Agribot projects, where I was the developer of vision and robot localization systems.Developed a proof-of-concept of tomato peduncle detection system that requires millimeter-level accuracy.𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 𝐬𝐭𝐚𝐜𝐤:- Python, YOLOv5, Detectron2, Norfair tracker, C++, TensorRT, AWS, ROS2𝐌𝐚𝐢𝐧 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧𝐬:- Implemented 2D object/keypoint detection and tracking algorithms- Implemented 2D/3D association algorithms based on stereo camera- Established in-house labeling team- Ran model training experiments on AWS cloud- Deployed TensorRT models on Jetson Nano and PCDeveloped obstacle detection system for autonomous orchard tractor.𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 𝐬𝐭𝐚𝐜𝐤:- C++, Python, PointCloudLibrary (PCL), ROS𝐌𝐚𝐢𝐧 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧𝐬:- Designed pointcloud-based obstacle and orchard row detection pipeline- Implemented pointcloud processing pipelines for lidar and stereo camera- Ran field tests with lidars, depth cameras and ultrasound sensorsDeveloped centimeter-level localization system for autonomous orchard tractor.𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 𝐬𝐭𝐚𝐜𝐤:- ROS, C++, Python𝐌𝐚𝐢𝐧 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧𝐬:- Ran field tests with RTK GNSS receivers, IMUs and wheel encoders- Developed ROS-based sensor and localzation simulations- Tuned localization Kalman Filter with data from field tests and simulations

- Administered computer systems and maintained mobile robots