Develop and implement AI/ML models for medical and natural images
Requirements
- Solid programming skills in Python and computer vision fundamentals; OpenCV desirable;
- Practical knowledge of TensorFlow or PyTorch for training and inference;
- Basic knowledge of C++ and/or Java to integrate inference into production (e.g., via ONNX Runtime); familiarity with CUDA/TensorRT is a plus;
- Knowledge of Git/GitHub, SQL, and Excel for organizing data and results;
- Experience with MONAI and nnU-Net in 2D/3D segmentation.
- Experience with conversion/optimization (ONNX, FP16/INT8 quantization) and GPU performance instrumentation (TensorRT).
- Practical experience with LLMs (fine-tuning, serving, Transformers) and orchestration (LangChain/RAG).
Responsibilities
- Develop, train, and validate AI/ML models (classification, detection, and segmentation) for medical images and natural images;
- Build data pipelines (preprocessing, normalization, registration, augmentation) in Python using PyTorch/TensorFlow;
- Explore and adapt reference frameworks for medical and general segmentation, such as nnU-Net (training, post-processing, and evaluation);
- Convert and optimize models (PyTorch/TensorFlow → ONNX) and implement inference with ONNX Runtime integrated with C++/Java applications and Python services;
- Accelerate inference on NVIDIA GPUs via TensorRT (FP16/INT8) when applicable to production;
- Apply image processing concepts (filters, morphology, equalization, registration) in pre/post-processing;
- Develop interfaces and support tools (visualization/labeler) in JavaFX (FXML) and/or C++/Qt;
Other
- Bachelor's degree in Engineering, Computer Science or Related Area;
- Collaborate with software teams for packaging and monitoring models.
- Travel requirements not mentioned
