Lead ML Engineer - Lane & Route Network Mapping

May Mobility is transforming cities through autonomous technology to create a safer, greener, more accessible world. They are seeking a Lead ML Engineer to architect the next generation of their mapping and localization stack, focusing on building production-grade semantic and topological foundations for autonomous vehicles. The role involves leading the design and implementation of advanced neural networks for lane and route network mapping, ensuring high-performance integration with various modules.

Responsibilities:

• Lead the research, design, architecture, training and validation of advanced neural networks for vectorized mapping (e.g., MapTR), multi-camera BEV transformers, and multimodal fusion models to extract and model lane and route networks for both high-fidelity offline pipelines and real-time online mapping
• Architect, design, and implement a production-grade lane and route network mapping stack, ensuring high-performance integration with upstream and downstream modules like Perception, Behavior, Policy, and Prediction
• Drive major feature development from inception to deployment. This includes high-level architecture design, rigorous code reviews, automated testing, mentorship of junior engineers, and technical resolution
• Own the end-to-end data strategy for the mapping domain, specifically focusing on lane and route networks. You will define data curation, auto-labeling, synthetic data, and active learning pipelines to capture and resolve long-tail scenarios
• Develop robust metrics and evaluation frameworks for lane and route network accuracy, temporal consistency, and scaling across diverse Operational Design Domains (ODDs)
• Work independently with cross-functional teams to translate complex autonomy goals into clear software and system requirements
• Collaborate with ML and Autonomy engineers to ensure the seamless deployment and validation of mapping features to the vehicle fleet
• Stay at the research frontier by evaluating, adapting, and innovating cutting-edge techniques, including online vectorized HD map construction, end-to-end mapping models, and vision/fusion Foundation Models to deliver production-ready solutions

Requirements:

• Ph.D. or Master's degree in Computer Science, Electrical Engineering, Robotics, or a related field with a strong mathematical and engineering foundation
• 7+ years of industry experience developing and deploying ML/DL models for mapping or computer vision at scale
• Deep expertise in several of the following areas: Vectorized mapping networks (e.g., MapTR), BEV-based scene representation, and temporal modeling
• Cross-modal calibration and fusion (e.g., Camera-to-LiDAR) within Bird's-Eye-View (BEV) unified representation spaces
• Transformers or Graph Neural Networks (GNNs) applied to structured lane geometry and topological connectivity
• Lane-level topology and connectivity, intersection modeling, and lane/road network graph construction
• Computer Vision Foundations: Object detection, classification, segmentation, tracking, depth estimation, and 3D reconstruction
• Strong understanding of HD maps, including lane and road network geometry modeling, connectivity, and semantic attributes
• Expertise in ML/DL development using PyTorch or TensorFlow, including experience with distributed training, synthetic data generation, large-scale dataset handling, and data curation strategies
• Strong programming skills in Python and/or C++ with experience in modular software design and Linux-based development
• Proven leadership in guiding technical roadmaps, mentoring engineers, and driving measurable improvements in model performance and system reliability
• Strong communication skills with the ability to lead technical discussions and align with cross-functional teams
• 10+ years of experience in ML/DL for autonomous driving or ADAS systems
• Experience with self-supervised and/or semi-supervised learning for large-scale representation learning
• Experience utilizing Vision-Language Models (VLMs) and/or Foundation Models for auto-labeling and long-tail (edge-case) detection
• Expertise in ML optimization for real-time products with limited compute, such as quantization, pruning, or distillation of large transformer models
• A proven record of inventions and/or publication record at top-tier conferences (e.g., CVPR, NeurIPS, ICCV, ECCV, ICLR)