Machine Learning Engineer, Physical AI
We're building the data engine behind Physical AI — the perception and spatial-reasoning systems that let robots understand the real world. As our Machine Learning Engineer, you'll own the pipelines and models that turn raw multi-sensor data into training-grade datasets, and fine-tune the computer vision models that depend on them.
We're a high-growth company: you'll work directly with the founders and with many world-tier robotics companies, in a hands-on role that spans training detectors, writing data-cleaning algorithms, and reasoning about geometry in a SLAM stack.
What you'll do
- Fine-tune and optimize computer vision models (YOLO and similar detection/segmentation architectures) for real-world Physical AI tasks, and iterate on them based on data and failure analysis.
- Design and build data-cleaning algorithms and pipelines — deduplication, outlier and mislabel detection, automated QA checks, active-learning loops, and label-consistency tooling — to produce high-quality datasets at scale.
- Develop SLAM-related applications and tooling: ingest and validate camera/LiDAR/IMU data, support mapping and localization workflows, and surface data issues that degrade spatial accuracy.
- Define and track data-quality metrics, build dashboards and validation gates, and root-cause quality regressions across the pipeline.
- Work closely with perception, robotics, and ML teams to translate model failures into concrete data improvements.
What we're looking for
- 3–6+ years of experience in computer vision, machine learning, or data engineering, with a track record of shipping work into production.
- Hands-on experience training and fine-tuning CV models (YOLO, Faster R-CNN, SAM, or similar) and solid fundamentals in CNNs, object detection, and segmentation.
- Experience building large-scale data pipelines and datasets, and writing algorithms for data cleaning, validation, or quality assurance.
- Working knowledge of SLAM concepts and frameworks (e.g., ORB-SLAM, RTAB-Map), multi-sensor data (cameras, IMUs, LiDAR), and sensor fusion or calibration.
- Strong attention to detail, a quantitative mindset about data quality, and the ability to collaborate across perception and robotics teams.
Nice to have
- Experience with 3D computer vision or probabilistic state estimation (Kalman filtering, pose estimation).
- Model optimization for deployment (ONNX, TensorRT, quantization/pruning) and GPU/cloud training infrastructure.
- Experience with active learning, auto-labeling, or human-in-the-loop annotation systems.
- Background in robotics, autonomous vehicles, AR/VR, or other Physical AI domains.
Compensation
- Competitive cash base + performance-based bonus + equity.
Why join
You'll work in person in San Francisco alongside a team building foundational infrastructure for embodied intelligence, with direct ownership over the data quality that determines how well our models perceive and navigate the physical world.