Member of Technical Staff, Computer Vision (Perception)

present of AI; San Francisco (CA); On-site; USD 150,000 - 190,000; Full time

Job summary

present of AI in San Francisco seeks a Member of Technical Staff, Computer Vision (Perception) to drive hands-on CV in products like robots and autonomous systems. You will work on camera geometry, SLAM, and 3D reconstruction, writing solid Python with PyTorch and shaping data into production-ready pipelines.

You’ll translate customer needs into robust training data, design fair tests, and collaborate with frontier labs on large video collections to influence model quality.

Qualifications

• Hands-on computer vision experience from real products (robots, headsets, or self-driving systems).
• Minimum of three years of experience.
• Knowledge of camera geometry, SLAM, and 3D reconstruction.
• Proficiency in Python with PyTorch.
• Ability to translate data into training data and production pipelines.
• Comfort talking to customers and taking a pipeline to production.

Responsibilities

• Build data for training AI models and evaluate data usefulness.
• Design fair tests and catch subtle errors.
• Turn customer needs into production-ready data pipelines.

Skills

Camera geometry

SLAM

3D reconstruction

Python

PyTorch

Data quality assessment

Customer communication

Production pipeline

Tools

PyTorch

Job description

Member of Technical Staff, Computer Vision (Perception)

The role asks for hands-on computer vision experience from real products such as robots, headsets, or self-driving systems, with at least three years behind it. A strong candidate knows camera geometry, SLAM, and 3D reconstruction, writes solid Python with PyTorch, and can handle messy multi-sensor data. Just as important is good judgment about what makes data useful for training, plus comfort talking to customers and taking a pipeline all the way to production.

How it fits the AI era

Much of the work is building the data that other AI models learn from, so the job sits right at the base of the AI stack. Models can now help with tasks like tracking, segmentation, and spotting bad footage, which lets a small team process far more video. What stays human is deciding what good data looks like, designing fair tests, catching subtle errors, and turning fuzzy customer needs into something that works.

How it builds AI experience

The work involves fine-tuning vision and vision-language models and building evaluations that show whether a dataset helps robotics and world models. Working with frontier labs on large video collections gives a close view of how training data shapes model quality. Someone wanting to build toward this could practice on public egocentric video sets, recover camera paths from footage, and write up how they measured data quality.

Worth showing

• SLAM and structure-from-motion
• Multi-sensor calibration and sync
• Benchmarks for robotics data

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Sieve

Multimodal artificial intelligence data laboratory creating curated video, audio, and physical world training environments for frontier foundation models, per the company.