Perception Research Scientist

TuSimple is a global autonomous driving technology company headquartered in San Diego, California, with operations in the United States and Asia. Founded in 2015, TuSimple is developing a commercial‑ready, fully autonomous (SAE Level
4) driving solution for long‑haul heavy‑duty trucks. TuSimple aims to transform the $4 trillion global truck freight industry through the company's leading AI technology, which makes it possible for trucks to drive safely autonomously, operate nearly continuously, and reduce fuel consumption by 10%+ relative to manually driven trucks. Global achievements include over 200 Patents, the world’s first fully autonomous, “driver‑out” semi‑truck run on open public roads in the U.S. and China, and development of the world’s first Autonomous Freight Network (AFN).

TuSimple’s Perception Team develops cutting‑edge perception technologies to help our trucks perceive the world. Perception Research Scientists are the “theory experts” who delve deep into the underlying principles and theories behind technologies. They will play an important role in creating novel algorithms to solve the most challenging perception problems and apply your solution to the fleet of trucks y will also work closely with other talents in this field in building the next generation of autonomous sensing algorithms.

They often aim to advance the fundamental knowledge of the field. TuSimple’s Perception Research Team are the “thinkers” who explore new ideas and methods. They work on creating new techniques or improving existing ones to make the autonomous driving system smarter and more efficient.

• 2D/3D perception including detection, segmentation, BEV‑related 3D scene understanding, etc., in challenging real‑world scenarios.
• Multi‑sensor fusion, tracking, and state estimation using camera, LiDAR, RADAR, etc.

• Develop cutting‑edge perception algorithms for autonomous driving trucks, using camera, LiDAR, RADAR, etc.
• Research deep learning and machine learning for the most challenging perception problems, including 2D/3D perception, sensor fusion, tracking, state estimation, etc.
• Deliver high‑quality, robust, and reliable code for the perception system on the L4 autonomous driving trucks.
• Explore new algorithms or improve existing ones to advance the technology, often publishing findings or implementing them into the existing system.
• Focus on conceptual work, sometimes without immediate practical applications. Pose questions, develop theories, and perform experiments to answer those questions. The goal is often to expand the foundational understanding of a subject.
• Use statistical software for data analysis like R or specialized scientific computing platforms like MATLAB. Often publish findings in academic journals, using LaTeX for documentation.
• Research and prototype development using deep learning and machine learning algorithms for the most challenging problems in multi‑sensor perception and fusion, including camera, LiDAR, RADAR, etc.
• Research is often done using high‑level programming languages like Python and data science libraries. May also use scientific computing platforms and specialized hardware for experimentation.

• Master’s Degree is required for this role or at least 2 years of professional experience in applying deep learning, machine learning, or computer vision algorithms on a large scale and real‑world data.
• Experience and strong knowledge in at least one of the following:
  • Camera‑based 3D detection or depth estimation
  • Multi‑sensor multi‑modal association/fusion
  • RADAR Perception
  • Occupancy Grid
  • Scene Understanding
  • Production‑quality coding skills in Python and/or C++.

• Master’s Degree specifically in Computer Science, Electrical Engineering, or related field.
• Ph.D. in Computer Science or Electrical Engineering.
• Prior academic or industrial experience in working with autonomous vehicles or perception systems.
• Track record of publishing in top‑tier computer vision or machine learning conferences and/or journals.
• Experience in designing and deploying algorithms for real‑world safety‑critical systems.