居家办公 Staff Machine Learning Engineer: Computer Vision 工作, 薪水, Arcus Ai Limited 公司招聘中 - Ricebowl

About Arcus

Arcus is the end-to-end platform for fire alarm design, and the automation backbone for the broader future of building-systems engineering. We help MEP (mechanical, electrical, plumbing) engineering firms automate the compliance-heavy parts of their work: reading architectural floor plans, placing system components at code-compliant locations, generating takeoffs, and producing project documentation. Work that used to take days now takes seconds.

We're a small, ambitious team building serious technology for a sector that has been waiting decades for it.

The role

We're hiring our first dedicated Machine Learning Engineer to own the computer vision system at the heart of the product. Floor plans are how engineers communicate; our ability to read them accurately, automatically, and to a code-compliant standard is what makes Arcus possible. You'll own the model, the dataset that trains it, and the evaluation methodology that keeps both honest.

This is a staff-level role with meaningful ownership. You'll define the ML roadmap, make the build-vs-buy calls, and shape the team that comes next as we expand from fire alarms into the full stack of building systems.

What you'll do

• Lead the development of our computer vision system for understanding architectural drawings: rooms, walls, doors, windows, and the fixtures and layout structure that downstream automation depends on.
• Build our proprietary floor-plan dataset from the ground up: sourcing, annotation strategy, quality assurance, versioning, and licensing.
• Define and own the evaluation methodology (pixel, polygon, and instance-level metrics) and the regression harness that gates every model release.
• Decide where foundation models and LLMs stay in the loop, where dedicated trained heads take over, and where classical computer vision is the right answer.
• Ship models into production: train, evaluate, package, deploy, and monitor. Partner with engineering and product to turn AI capability into customer-visible automation.
• Hire and mentor the ML engineers who join after you.

What you'll bring

• 7+ years of production machine learning experience, with deep focus on computer vision.
• A track record shipping semantic or instance segmentation models in real products.
• Strong Python and PyTorch (or equivalent framework) fluency.
• Hands-on experience building annotation and dataset pipelines from scratch, not just consuming someone else's.
• Disciplined approach to evaluation: you define the metrics before you train, ship eval harnesses alongside models, and gate releases on numbers.
• Comfort with model deployment, MLOps, and serving models behind production APIs.
• The judgment of a senior IC: you know when to train from scratch, when to fine-tune, when to use a foundation model, and when classical computer vision is the right call.

Bonus points

• Experience with document, diagram, CAD, or layout understanding: architectural drawings, technical schematics, scanned forms.
• Familiarity with public floor-plan or layout datasets (CubiCasa5K, R3D, PubLayNet and similar) and their licensing realities.
• LLM-in-the-loop systems for labelling, bootstrapping, or hybrid inference.
• Active learning, weak supervision, or synthetic data generation.
• Exposure to CAD / BIM formats (DXF, IFC, SVG) or the broader BIM ecosystem.
• Prior work in spatial AI, proptech, construction tech, architectural automation, or document AI.

Why Arcus

• Real product, real customers. We're already in production with engineering firms, not pitching from a deck.
• First ML hire. You set the technical standard and lead the team that follows.
• A real moat. You'll build the dataset and the model that competitors will spend years trying to catch up to.
• A big sector. MEP engineering is a multi-billion-pound industry that has barely begun its software transition.
• Fully remote. Globally distributed, async-friendly.

How to apply

Send your CV and a short note on a recent machine learning project you're proud of (what it does, what you owned, and what you'd do differently a second time) to *************