Principal Machine Learning Engineer

Who We're Looking For

As a Principal Machine Learning Engineer in Delivery, you are an experienced problem solver and technical leader who stays anchored to impact. You can grasp advanced engineering concepts across multiple industries, lead technical initiatives, and work directly with customers (often side by side on-site) to embed cutting-edge AI models into tools that are useful and used.

For this role, that means manufacturing and maintenance, repair and overhaul (MRO). You have put ML into production on industrial process or sensor data, not benchmarks. You know the difference between a model that scores well offline and one a process owner will accept on a qualified production line. Your data comes off equipment rather than out of a solver: it is sparse, drifting, miscalibrated and partly unrecorded, and you know how to work with it.

You've shipped ML systems end to end and at scale: you design, build and test reliable, scalable ML data pipelines, select the right libraries, frameworks and tools, and make pragmatic product decisions that set Delivery up for success. Working at the intersection of data science and software engineering, you translate R&D and project outputs into reusable libraries, tooling and products.

With at least 5 years of industry experience (post Masters or PhD) in a commercial, non-research environment, you're ready not only to execute but to lead and mentor others.
This Role

You'll work closely with Data Scientists, Simulation Engineers, Manufacturing Process Experts and customers to define the process and physics problems we are solving, and you'll drive decisions on reliable deployment with measurable outcomes on the shop floor.

You'll:

• Own the ML approach to manufacturing and MRO data, from modelling through deployment in customer production environments (cloud, on-prem and edge).
• Build production ML on sparse, drifting, miscalibrated process and sensor data (process logs, sensor traces, inspection records, small DoE sets), and judge whether the data is enough to train on.
• Choose between physics simulation, reduced-order and learned models, and fuse simulation with shop-floor measurement where the two disagree.
• Set the UQ and validation standard for what a process recommendation may claim against a qualified process window, and design experiments where machine time is the cost.
• Make models transfer across machines, material lots and consumable batches.
• Define the product requirements that manufacturing and MRO place on the PhysicsX platform.
• Lead scoping and architecture for data and ML systems, and communicate results and trade-offs to senior stakeholders and process owners.
• Build end-to-end training and inference pipelines and APIs (Python, PyTorch, Pandas, FastAPI, SciPy, Kubeflow, among others), including for 3D point-cloud and mesh data, and mentor engineers and data scientists.
• Travel to customer sites worldwide, on average 3-4 weeks per quarter, and take part in scoping new projects and bringing new customers to PhysicsX.

What you've done

We hire on what you've done, so strong candidates can point to things like:

• Built and deployed ML on manufacturing process data (parameter logs, sensor traces, inspection records or small DoE sets) and can say what changed in production as a result
• Fused simulation output with shop-floor measurement where the two disagreed
• Set UQ and validation standards that other teams followed
• Shaped a product or tooling roadmap based on what a site or company actually needed
• Had a recommendation rejected by a process owner, and understood why

Experience in aerospace or another regulated manufacturing environment is a plus. It is not required, because the must-have bar can be met in any sector where data comes off equipment.

Salary range:

$140,000 - $190,000 depending on experience
Seniority will be assessed throughout our interview process