Job Overview
We are seeking a skilled AI/Machine Learning Researcher to join our team. The successful candidate will be responsible for developing and enhancing video analytics capabilities, focusing on improving accuracy, robustness, and efficiency of current models.
Key Responsibilities:
* Investigate, benchmark, and optimize model architectures for detection, segmentation, and tracking in complex transport environments.
* Enhance multi-object tracking pipelines using advanced methods such as DeepSORT, ByteTrack, and Transformer-based approaches.
* Improve video analytics performance through evaluation of alternative models and pipeline optimization.
* Develop post-processing logic for trajectory denoising, smoothing, and error reduction.
* Apply denoising techniques to improve low-resolution or low-bitrate footage.
* Conduct field validation of model performance across diverse environments and camera types.
* Support camera calibration workflows for accurate spatial measurements.
* Run ablation studies and model evaluations under varied environmental conditions.
* Collaborate with cloud engineering teams for scalable inference deployment using ONNX, TensorRT, Triton, or similar.
* Translate AI outputs into transport safety and engineering metrics.
* Contribute to documentation, reports, and publications.
Essential Criteria:
* A PhD in Computer Science, Applied Mathematics, Robotics, Physics, or related discipline; or equivalent experience with a strong applied AI/computer vision track record.
* Expertise in object detection, segmentation, and multi-object tracking in complex scenarios.
* Proficiency with deep learning frameworks (PyTorch, TensorFlow) and Python libraries (OpenCV, Yolo, NumPy, scikit-learn).
* Proven experience in real-world video analytics and model deployment under challenging conditions.
* Experience with model optimization and inference acceleration.
* Familiarity with denoising techniques for low-quality video.
* Understanding of camera calibration and spatial measurement.
* Strong communication skills and ability to work independently and in multi-disciplinary teams.
Desirable:
* GPU-based inference deployment (ONNX, TensorRT, Triton).
* Knowledge of transport/traffic datasets.
* Track record of publications or open-source contributions.
* MLOps tools and practices (MLflow, Weights & Biases).
* Experience with Transformer architectures for video data.
At Paxus, we value diversity and welcome applications from people with diverse cultural backgrounds and living with a disability. If you require adjustments to the recruitment process, please contact us.
With Paxus, you will have the opportunity to work on impactful projects at the intersection of cutting-edge technology and transportation safety.