Senior Deep Learning Model Engineer

What we're building

OrbDB is building data infrastructure for AI reliability. For every prediction a model makes, the platform determines whether the model is sufficiently certain for the result to be acted on automatically or whether the case should be routed to a human reviewer. 

Today’s AI production systems are unable to distinguish which of their predictions are trustworthy. We are building the layer that allows organizations to automate the cases where automation is statistically justified, and to escalate the rest with confidence.

OrbDB is founded and led by researchers with deep expertise in the underlying methods.

The role

You will work on the models that sit at the center of our platform. Our work is built around Graph Neural Networks, and the questions you will engage with are the ones that sit beneath the surface of any serious deep learning system: questions about architecture, training behaviour, optimization, and the relationship between what a model is doing and what we expect it to do.

This is a role for someone who knows the fundamentals of deep learning well enough to reason about them from first principles, not from tutorials. You will work closely with our research-led founding team, and the questions you take on will move between the practical and the foundational, often within the same week. Unlike other AI startups, OrbDB builds on a mathematical foundation. So do the teams behind it. OrbDB Labs is a place where solid ideas and good taste matter more than loud voices.

Specifically, you will:

• Train, evaluate, and improve the models that power the platform.
• Diagnose model behavior at a level deeper than metrics, and propose changes grounded in the underlying mathematics.
• Make principled choices about model design as required.
• Work alongside the engineering team to deliver research-grade models into a production system that customers can rely on.

What we are looking for

• 2-4 years of experience working with deep learning models in a serious technical setting, whether in research, industry, or a combination. If you are close to that range and the rest of the role fits, we would still like to hear from you.
• A real command of the fundamentals of deep learning. You should be comfortable reading a paper, implementing it, and reasoning about why a model is or is not behaving as expected.
• Strong engineering skills. You write code that others can build on, and you understand that a model is only useful once it runs reliably.
• Fluency with the modern deep learning toolchain, particularly PyTorch.
• Genuine interest in the statistical foundations of what we are building. Concepts like Conformal Prediction and calibration should be ones you are eager to understand deeply.

Discover our full range of relocation jobs with comprehensive support packages to help you relocate and settle in your new location.

Useful, but not required

• Experience with Graph Neural Networks specifically, or with the libraries that support them (PyTorch Geometric, DGL, or equivalent).
• A graduate degree in a quantitatively rigorous field, or equivalent depth acquired through other means.
• Open-source contributions in the ML or deep learning ecosystem, particularly to production-grade libraries.
• Experience moving models from research code into production systems.

How to apply

You can read more about our work at orbdb.io. To apply, please send your CV or LinkedIn profile to contact@orbdb.io, together with a short note about a system you have built that you are proud of, and what you would approach differently.

This is a Stockholm-based hybrid role. We expect the candidate to be located in the Stockholm area. If you are currently elsewhere in Europe and are interested in moving, we are happy to discuss relocation support.