POINT.LAZ - Director of AI (Artificial Intelligence)

The Mission

The Lazaruss scanner fuses LiDAR and high-resolution imaging to record the geometry and surface condition of vertical mine shafts — buntons, guides, compartments, skips. Your job is to make those detections trustworthy enough that a shaft inspector relies on them to generate work orders that cost money and affect safety. The gap between “the model works on our dataset” and “the model works at 1,400 m depth, covered in dust, on a Thursday morning in Mongolia” is the gap you will close. We are deployed on three continents with a global distribution partner. The product exists. The algorithm is in development. What we need is a Product Owner who knows how to move from it works to it ships, it is trusted, and it gets better every release.

What You Will Own

The full path from algorithm to product surface. You define what “good” looks like in production — not F1 on a curated benchmark, but the operational metrics the mine actually pays for: false-positive burden on the reviewer, time-to-first-detection, depth coverage, bunton-state agreement with the human expert. You translate inspection workflows — filtering by depth, validating detections, generating a work request, comparing against the previous scan — into a product roadmap that the engineering team can ship against. You partner with Software Development to land models inside the Visualizer, with Hardware so inference constraints are designed in (not bolted on), and with Quality to make outputs defensible under the regulatory and liability weight the mining industry carries.

A Particular Emphasis

Moving from research-grade code to MLOps-grade systems — versioned datasets, reproducible training pipelines, continuous evaluation against production data, monitored drift, and a disciplined release process. If you have done this transition once and have the scars, you will stand out.

You have done this before - We are hiring for one specific thing: a completed trip from prototype to deployed product.

• •Experience taking a computer-vision or 3D-perception system from internal prototype to external, paying customers. This is the game-changer. Everything else can be taught.
• •3+ years as a Product Owner or equivalent on AI/ML products - You can read a model architecture diagram and push back on it, without writing the model yourself.
• •Literacy in modern deep learning - CNNs, vision transformers, point-cloud methods (PointNet / PointNet++, voxel and sparse-convolution approaches), segmentation and detection pipelines, classical CV
• •Hands-on MLOps fluency - Dataset curation, model versioning, CI/CD for ML, shadow deployments, A/B testing, drift monitoring
• •Fluency in French or English - Our team is in Québec. Our customers are on three continents

Nice to Have

• •Edge deployment on hardware like NVIDIA Jetson, running in dust, vibration, and humidity - Industrial inspection, NDT, robotics, autonomous vehicles, or medical imaging backgrounds all transfer
• •Uncertainty quantification, out-of-distribution detection, active learning - Safety-critical outputs need to say not only what the model concludes, but how much to trust it this shaft, this day
• •Familiarity with mining, geotechnical engineering, or structural inspection - We can teach the mining; we cannot easily teach the product instincts
• •Exposure to ISO 9001, functional safety concepts, or equivalent industrial certification regimes

Twelve Months In - What Success Looks Like

• •A Lazaruss release is out where AI-driven diagnostics are a headline feature used by paying customers.
• •Measurable reductions in inspector time-per-shaft. A dataset and evaluation pipeline that lets us ship improvements on a predictable cadence. Customer feedback loops that will let us scale the capability across the rest of the product. The philosophy of AI-assisted inspection — augmenting the expert rather than replacing them — is visible in every part of the UX.

APPLY - Send us the one that didn’t go smoothly

• •We don’t want a resume and a cover letter. We want your CV and a short note describing one AI product you took from concept to deployment — what worked, what did not, and what you would do differently. Specifics beat polish.