Software Engineer - Perception

Lodestar•San Francisco, CA

14d•$115,000 - $180,000

About The Position

As a Software Engineer - Perception at Lodestar, you’ll be joining the perception and machine vision teams to rapidly grow our vision and perception models at MITHRIL’s center. You’ll focus on researching and building algorithms to detect, classify and understand unknown space targets in real-time. By combining computer vision, image processing and machine learning, you’ll develop custom models - providing the foundation for autonomous decision-making in complex and uncertain space environments. We proudly have an "extreme ownership" oriented engineering culture.

Requirements

Bachelor’s or Master’s degree in, AI/ML, Computer Science, Robotics, a related field, or equivalent experience
2+ years of distinguished industry experience in Machine Learning, robotics, computer vision, or a related field
Excellent software engineering skills in C++ and Python
Strong DNN/ML background and experience in CV and image processing libraries (OpenCV, CUDA-accelerated vision, PCL)
Broad understanding of modern perception systems and their architecture
Experience developing and deploying ML models to real-time systems

Nice To Haves

Experience processing sensor measurements from a variety of sensors including RADAR, electro-optical sensors, RF sensors, or LIDAR
Familiarity with image processing techniques (e.g., feature extraction, segmentation, filtering, stereo matching) to support 3D reconstruction and target classification
Track record optimizing neural networks and geometric algorithms using CUDA kernels, TensorRT, or other GPU acceleration frameworks
Experience with distributed training and cloud-based scaling of ML models (AWS, GCP, or Azure)
Experience with Linux, Git, and CI/CD pipelines
Comfortable with containerization tools such as Docker and Kubernetes
Strong understanding of networking and communication protocols for distributed autonomy systems

Responsibilities

Scoping architecture for the perception system core to Lodestar’s autonomy software suite
Research and adapt novel perception algorithms - everything from lit. review, through training and evaluation, to optimisation and deployment
Design, develop ML algorithms from the ground up, in addition to build on and optimizing our existing models for edge computing
Further develop our robust methods for predicting relative pose of dynamic space targets in harsh and unpredictable visual environments
Integrate perception models into mission simulation environments and real-time autonomy pipelines
Contribute to the design of realistic sensor models and digital twins to improve perception fidelity in mission scenarios
Own perception tooling, model training infrastructure, and evaluation frameworks for ongoing development and deployment
Collaborate across autonomy, estimation, and mission software teams to ensure perception fidelity, performance, and scalability