Samsara is looking to improve the safety, efficiency, and sustainability of physical operations by developing and deploying efficient, reliable, and scalable AI models for their in-vehicle intelligence systems on constrained embedded environments.
Requirements
- 5+ years of experience developing and deploying deep learning models for edge, embedded, or real-time systems.
- Strong background in computer vision or multimodal ML (e.g., 2D/3D CNNs, Transformers) using industry-standard deep learning frameworks.
- Proficiency in Python and C++, with hands-on experience optimizing inference runtimes and applying model optimization techniques for edge deployment.
- Deep understanding of performance tuning, including compiler- or DSP-level optimizations, runtime profiling, latency analysis, and memory management on constrained hardware.
- Familiarity with middleware or streaming frameworks used in real-time perception pipelines.
- Experience bringing ML infrastructure or runtime systems from prototype to production at scale.
- Background in multimodal ML (e.g., audio + vision fusion) or event-based detection systems.
Responsibilities
- Design, optimize, and deploy computer vision and multimodal ML models that run efficiently on constrained edge platforms powering Samsara’s in-vehicle camera systems.
- Apply advanced model optimization techniques—such as quantization, pruning, and distillation—to achieve real-time inference under strict CPU, memory, and thermal constraints.
- Partner with ML research and product teams to translate new AI detections into deployable, maintainable edge models.
- Collaborate with firmware, ML research, and hardware teams to productize our ML runtime pipeline, bringing scalable, reliable, and testable on-device inference to production.
- Develop performance benchmarking, profiling, and validation frameworks for edge-deployed models to ensure robustness across millions of deployed devices.
- Drive continuous improvement of our edge ML toolchain and advocate for best practices in model optimization, inference reliability, and deployment efficiency.
- Mentor peers on efficient inference design and collaborate cross-functionally to accelerate feature delivery for safety and driver experience programs.
Other
- This is a remote position open to candidates residing in the US.
- You want to impact the industries that run our world.
- You want to build for scale.
- You are a life-long learner.
- You believe customers are more than a number.
