Staff Machine Learning Engineer

Summary

Join Stack's Static World Perception team as a Staff or Senior Staff Engineer and lead the development and improvement of algorithms and systems for online mapping in L4 self-driving vehicles. You will integrate various data sources (maps, LiDAR, cameras, GPS) to generate road elements governing traffic flow. Responsibilities include contributing to the team's technical direction, writing robust software for real-time applications, driving cross-functional features, identifying system limitations, and aligning stakeholders. The ideal candidate possesses extensive experience in deep learning model deployment, applied research, perception/mapping solutions for robotics, and strong software engineering and machine learning skills. Fluency in Python and experience with C++ are required.

Requirements

• Extensive experience architecting, training, and deploying deep learning models into real-world environments
• Track record of driving applied research projects from start to completion, including conception, problem definition, experimentation, iteration, and finally publication or productization
• Experience delivering perception/mapping solutions for real-time robotic applications
• Strong experience in software engineering and machine learning algorithm design
• Fluency in Python
• Experience with C++

Responsibilities

• Work in the team to deliver state-of-the-art online and offline mapping algorithms and systems for L4 self-driving vehicles
• Contribute to the technical direction of the team and broader perception org
• Write robust software for real-time, resource-constrained, safety-critical applications
• Drive highly cross-functional features and work streams from problem statement to delivery
• Proactively identify limitations with the existing online mapping system and drive improvements
• Align stakeholders with strong presentation and communication skills
• Lead challenging technical topics and build consensus

Preferred Qualifications

Prior experience in sensor fusion and/or online mapping is highly desirable