The role
- COMP
- $180K - $245K
- EQUITY
- Competitive equity
- LOCATION
- San Francisco
- WORKPLACE
- On-site
- EXPERIENCE
- 4+ years
- VISA
- None, Visa transfers
- STACK
- Python, PyTorch, OpenCV
- INDUSTRY
- AI, Environment, Manufacturing, Hardware, B2B
/roles — ROLE_465
Computer Vision ML / Research Engineer
Well-funded recycling tech company using stereo vision and ML to measure industrial material piles
The company
Recycling technology startup that sorts mixed scrap metal back into pure alloys and sells computer vision tools that help scrap yards value incoming loads and catch fraud.
- STAGE
- growth-stage
- FUNDING
- $40M+ raised
- TEAM
- ~10 people
- FOUNDED
- 2025
JD — the work
About the role
You would be responsible for the vision system that converts stereo images into precise, auditable volumes of material piles for industrial inventory. It combines stereo cameras and depth estimation with segmentation, point clouds, and calibration data to estimate 3D volume, and it is central to the product. Your remit runs from model integration and reconstruction quality to confidence metrics, taking the pipeline from research prototype to measurements customers can trust in production.
What you'll do
- Build the main CV pipeline covering reconstruction from stereo, segmentation, cleaning point clouds, fitting ground planes, stitching, and volume math
- Integrate modern stereo-depth and YOLO-style detection models
- Improve 3D reconstruction quality and volume accuracy, and test calibration assumptions
- Analyze failure modes and build confidence metrics and model evaluation
- Debug real-world sensor data captured at industrial sites
- Carry the pipeline from research prototype into a reliable production measurement system
What they're looking for
- 4+ years of experience in computer vision or ML engineering
- Strong Python and PyTorch, plus OpenCV or Open3D
- Solid skills in camera calibration, 3D geometry, and stereo or depth estimation
- Experience with point clouds and segmentation
- Comfort debugging messy real-world sensor data
Nice to have
- CUDA or experience with GPU cloud platforms
- Background in robotics perception, SLAM, or photogrammetry
- Industrial inspection or geospatial and volumetric measurement