Lead Perception Engineer
In-person · Los Angeles, CA, US... more
In-person · Los Angeles, CA, US... more
Job Description
We are Universe Energy, and we are the battery dismantling and sorting company.
The world needs 2 billion batteries by 2050 to make electric cars and power the grids. But we need to mine 30x more, leading to a $10 Trillion supply gap. We help EV OEMs, battery makers, and service centers make $1500 on a used battery pack by remanufacturing and repurposing them instead of letting them pay $750 for recycling. We dismantle, inspect, and grade used battery packs 90% cheaper and 6x faster using robotics, computer vision, and machine learning. We will deploy a fleet of 50 dismantling & sorting systems in shipping containers on-site to process 200k battery packs/year by 2030. Our mission is to unlock reuse as the primary source of the next 100 million batteries by 2050, powering a truly clean energy transition. We keep 5 TWh in battery capacity alive, reduce 6 Gigatons CO2, and save 125 billion liters of freshwater per year.
Our robot dismantles batteries.
We are building a cognitive robot that automatically diagnoses, discharges, and disassembles EV batteries using robotic manipulation, autonomous controls, and computer vision. The first-generation robotic system will automatically assess the battery’s state of health, remove covers from arbitrary battery packs (500 kg), and perform safe discharging. It then disassembles these batteries from the pack level (500 kg) down to the module level (25 kg). This system can take batteries apart 4x faster and safer than a human at 6x the throughput, leading to 50% lower unit economics.
Job objective
You will conceptualize, architect, engineer, and deploy algorithms that allow the robot to see and recognize the configuration of EV battery packs. The perception system then tells the robot what it sees and identifies parts like connectors, welding seams, and modules. You will scan the battery packs and cells to decide on their health by analyzing images from battery cell material. Then, instruct the robot on how to take these apart. You will then validate the software with what the cameras and sensors on the robot perceive.
How you will contribute
- Develop production-level & robust computer vision modules for classification, counting, tracking, 3D reconstruction, camera calibration, and segmentation.
- Research and implement machine perception & visual understanding of battery systems to enable counting, detection, localization, and labeling.
- Recognize, analyze, and process images of scans of battery cells from non-intrusive methods such as X-ray, CT-scan, and ultrasound.
- Build perception software that integrates computer vision, sensor fusion, decision-making functions, and structures data-set generation.
- Develop and implement classical and learning-based computer vision on real-time platforms.
- Perform sensor selection for the camera perception system like RGB, infrared, and laser scan. Develop sensor-fusion and decision-making algorithms.
- Generate datasets for algorithm training from the real world and through synthetic methods. Build software & ML infrastructure for machine perception capabilities.
The skills & experience that you bring
- At least a B.Sc. in Computer Science, Applied Mathematics, Machine Learning, or a similar field.
- Academic background in Applied Mathematics, Machine Learning, classical Computer vision, Image recognition, and Perception systems.
- 3-5 yrs experience in developing software with strong skills in C/C++, Python, and Matlab-Simulink and have developed software from architecture to production-level code in software, machine learning, and perception environments.
- 3-5 yrs experience in developing ML tools in Torch/TensorFlow built and Classical computer vision algorithms in C++ and OpenCV.
- 3 yrs hands-on experience with optical, image sensor, or camera calibration and their associated computer vision principles to process this data.
- 3 yrs experience in generating, filtering, and augmenting large image datasets for computer vision.
- 3 yrs experience developing, training, and testing deep-learning-based algorithms for detection, counting, classification, segmentation, and tracking.
How to hit a home run
- A track record of relevant academic publications, patents, and/or open-source software in machine learning and/or computer vision.
- Hands-on experience processing rich sensor data from LIDAR, RADAR, and cameras in environments captured by autononous vehicles.
- Experience in 3D graphics, focusing on 3D geometry manipulation (Vis-Rep, B-rep geometry representations) & Game engine experience in Unity3D (C#) or Unreal Engine (C++).
- Hands-on experience with building autonomous and/or robotic systems is a plus.