Senior Machine Learning Engineer, Computer Vision

About the Team: 

Torc's Autonomy Applications software utilizes cutting-edge deep learning techniques to perceive the vehicle's environment, predict the movements of other vehicles, and execute accurate driving decisions. We are actively seeking a highly experienced senior machine learning engineer to join our scene modeling team. This is an exceptional opportunity for you to have a significant impact on the future of the autonomous vehicle industry by leveraging AI. 

As a Senior Engineer of the team, you are applying machine learning science in a production focused environment. You are using machine learning models in both a unimodal and multimodal context, to create a 3D representation of the driving scenario. Training, validation, data science, architectural design are your daily work. You are interested in understanding how your model performs in deployment, for what you collaborate closely with deployment focused teams. You mentor and guide more junior members of the team and are always interested in the newest trends in research, eager to translate scientific improvements into our production grade machine learning pipelines. 

What you'll be doing:

• Develop and Optimize Computer Vision Algorithms: Implementing monocular and stereo depth estimation algorithms; comprehending objects, lanes, obstacles, and weather conditions within the driving environment; enhance perception systems to process multi-modal sensor data (camera, LiDAR, radar) effectively; and utilizing data science techniques to analyze model performance, data distributions, and identify corner cases
• Contribute to BEV Self-Driving Architectures:  Design and implement deep learning models for object detection, semantic segmentation, and voxel grid occupancy in BEV frameworks; integrate BEV representations into end-to-end planning and control pipelines 
• Data Management and Processing: Develop efficient pipelines for large-scale data processing and annotation(pseudo-labeling) of sensor data (e.g., LiDAR point clouds, image frames); and implement data augmentation, synthetic data generation, and domain adaptation strategies to improve model robustness
• Model Deployment and Optimization:  Collaborating with conversion and deployment teams to ensure seamless integration; deploy machine learning models on edge devices, ensuring real-time performance and resource efficiency; and optimize inference pipelines for embedded and automotive-grade hardware platforms
• Cross-functional Collaboration:   Collaborate with robotics, software, and hardware engineering teams to ensure seamless integration of perception systems; and work with product and operations teams to define performance metrics and improve system reliability
• Research and Innovation:  Stay updated with the latest advancements in computer vision, BEV models, and autonomous driving technologies; translating scientific research into production-grade machine learning pipelines; and publish findings in top-tier conferences and journals (optional but encouraged)
• Leadership:  Contributing to the model development roadmap and providing strategic advice to technical leadership and mentoring and guiding junior team members to enhance their technical skills and career growth; and mentoring and guiding junior team members to enhance their technical skills and career growth 

What you need to succeed:

• Bachelor's degree in computer science, data science, artificial intelligence or related field with 6+ years of professional experience or a master's degree with 4+ years of experience 
• Scientific understanding of machine learning for 3D BEV space modeling 
• Versatile in training PyTorch machine learning models
• Experience with understanding data distributions and analyzing long tail distributions 
• Mastery of Python and PyTorch, with the ability to transition research level code to production and deployment ready standards 

Bonus points!

• PhD in machine learning or data science 
• Proficient in writing CUDA kernels and developing custom PyTorch operations
• Experience with relevant NVIDIA libraries and frameworks, such as CUBLAS, CuDNN, and NPP