Member of Technical Staff, Computer Vision (Perception)

Physical AI runs on data that captures how the real world looks, moves, and fits together in 3D. As a perception-focused Machine Learning Engineer at Sieve, you'll build the systems that turn raw egocentric, robot, and sensor footage into the training data that robotics and world-model teams depend on.

You'll bring the computer vision instincts of someone who has shipped perception for robots, devices, or autonomous vehicles, and apply them to datasets at a scale few teams ever touch. One week you might recover camera pose and scene geometry across thousands of hours of first-person video. The next, you might build a hand and object tracking pipeline, or design a benchmark that tells a lab whether their data actually improves manipulation policies.

About Sieve

Sieve is a multi-modal lab curating the world's highest-quality training datasets, spanning video, audio, images, text, and 3D. We combine exabyte-scale data infrastructure and novel multimodal understanding techniques that push the frontier of foundation models. Video alone makes up 80% of internet traffic, and across modalities, data has become the enabling medium powering creativity, communication, gaming, AR/VR, and robotics. Sieve exists to solve the biggest bottleneck in the growth of these applications: high-quality training data.

We partner with top AI labs and did $XXM last quarter alone, as a team of ~30 people. We also raised our Series A from Tier 1 firms such as Matrix Partners, Swift Ventures, Y Combinator, and AI Grant.

Sieve is one of the most capital-efficient teams in AI, with roughly 30 people serving the world's leading AI labs across every major data modality. You'll join early, own problems end-to-end, and watch your work ship directly into the models defining the frontier.

Responsibilities

• Build perception pipelines for depth, pose, tracking, segmentation, and 3D reconstruction over large video and multi-sensor datasets
• Apply SLAM, visual odometry, and structure-from-motion to recover camera trajectories and scene geometry at scale
• Fine-tune vision and vision-language models for embodied tasks like hand-object interaction, activity understanding, and scene understanding
• Develop automated QA and filtering systems that catch calibration errors, sync drift, motion blur, and other data defects before they reach customers
• Create benchmarks and evaluation frameworks that measure how much a dataset improves downstream robotics and world-model performance
• Work directly with frontier labs and robotics teams to turn vague data requirements into production pipelines
• Ship fast, measure results, and iterate based on how the data performs in the real world

Requirements

• 3+ years building production perception or computer vision systems for robotics, AR/VR devices, autonomous vehicles, or similar
• Hands-on experience with SLAM, visual-inertial odometry, multi-view geometry, or 3D reconstruction
• Strong Python skills and fluency with PyTorch; C++ is a plus
• Comfortable with multi-sensor data such as camera intrinsics and extrinsics, IMU, depth, and time synchronization
• Experience training, fine-tuning, or deploying modern deep learning models for vision
• Sharp intuition for evaluation, edge cases, and what makes data actually useful for training
• Comfortable owning a problem from the customer conversation to a deployed pipeline
• Excited about Physical AI and the data that will power it
• In person at our SF HQ

Preferred

• Experience with neural scene representations such as NeRFs or Gaussian splatting
• Background working with egocentric or teleoperation data
• Publications or open-source work in perception, 3D vision, or robotics

Benefits

• 401k + Full Health Insurance
• Breakfast, Lunch, and Dinner covered and your choice of snacks
• Ubers covered home

All roles at Sieve require you to be onsite in San Francisco 5 days per week.