The Opportunity
Manufacturing powers the global economy at $50T a year and it relies heavily on human dexterity and skill to produce the goods we rely on every day. Yet manufacturers have little visibility when issues arise at manual assembly stations, impacting productivity and quality.
Assembler AI is changing that.
We use computer vision and artificial intelligence to help manufacturers improve quality, reduce waste, increase throughput, and enable frontline employees to perform at their best.
Backed by Diagram Ventures, we're building a category-defining company at the intersection of AI, manufacturing, and operational excellence.
The role
We're hiring a founding computer vision engineer. You'll work directly with the Head of AI on the perception models behind our products, and take ownership of the datasets and evaluation work that make them perform in the field.This is an applied, hands-on role. Our hardest problems aren't about picking the newest architecture , they're about building the right training data, evaluating honestly enough to know what actually improved, and spending real time in the footage to understand what a model is getting wrong. If you like problems where the answer is in the data rather than the paper, you'll do well here.
What you'll do
- Train, evaluate, and improve vision models on real production video
- Own our training datasets: sourcing and selecting data, defining label schemas, and reviewing annotation quality
- Design evaluations that reflect real deployment conditions, and report results clearly and honestly
- Investigate model failures directly on footage and identify their root causes
- Work with our annotation team to target the data most likely to improve performance
- Run experiments end to end: from question, to training runs, to a result someone can act on
Requirements
- Bachelor's in computer science, engineering, mathematics, or a related field ( or equivalent practical experience). A Master's in computer vision, machine learning, or a related area is welcome but not required
- 2+ years of hands-on applied computer vision experience, ideally in industry or on systems deployed beyond a research environment
- You've trained and deployed a vision model on real-world data end to end, including the messy parts
- Strong Python and PyTorch
- Solid understanding of evaluation: train/test splits, precision/recall tradeoffs, and why a strong validation score doesn't always mean a good model in production
- Comfortable spending significant time reviewing video footage; understanding what the camera actually sees is a real part of this job
- Willing to say "I don't know" and "I think I got that wrong." We'd rather hear it early
- Professional working proficiency in French and English
- Based in Canada
Technical environment
Required:
- Git and a standard branch/review workflow
- Linux - comfortable working entirely over SSH on remote GPU machines: bash, background jobs, services, reading logs
- Cloud - hands-on experience with AWS and/or GCP, including compute, storage, and managing your own training environments
- PyTorch and CUDA - GPU-based training, checkpoint management, and experiment tracking.
- OpenCV and ffmpeg - decoding, cropping, and processing video at scale
- NumPy, pandas, scikit-learn, and standard annotation data formats
Nice to have:
- Experience with video and temporal models, not just single images
- Object tracking and multi-object association
- Model export and inference optimization
- Experience with annotation platforms and running an annotation workflow
- Exposure to manufacturing, robotics, or industrial inspection
Why join
You'll be the second person on the AI team, with direct ownership of a core part of the product and a very short path from your work to something running on a live production line.
Compensation & Benefits
- Competitive salary and stock options
- Competitive Health benefits
- Health and wellness spending account, from $500 to $1,000 annually
- Latest MacBook and modern sales tooling
- Opportunity to grow as the company scales
