Projects

Things I have built, from the fjord to the shop floor.

Three years of taking machine learning into a production .NET product at Sandvik Coromant, and the PhD before it on robots that decide where to measure next. Each card opens the story behind it.

2026 A fault detector the shop floor can trust A causal streaming detector for cutting-insert faults, taken from a Python study to shipped C#. It caught every labelled fault without a single false alarm.
2026 From labelled cuts to C# constants An Azure ML pipeline that trains candidate detectors in parallel, picks a champion against the challengers, and exports the winner straight into production code.
2025 Thousands of sensor traces, drawn on the GPU A legacy SVG chart re-architected into a WebGL2 and Canvas engine that overlays 1000+ series per plot and opens 14-hour recordings, with comparison tools built on top.
2025–26 Comparing a cut with a good one, while it happens A live view that streams machining data against a known-good reference curve, so an operator sees a deviation during the cut instead of in the report afterwards.
2025–26 Sharing recordings without sharing the keys Central, audited storage for sensor recordings on Azure: Blob Storage and a document database behind least-privilege roles, managed identity and automated delivery.
2024–26 One parser for two languages A C# gRPC service with a Python client wheel that opens production's binary recording format to notebooks and Azure ML, so analysis and product read data the same way.
2024 Deep models inside a .NET product LSTM, autoencoder and VAE models trained in PyTorch, exported to ONNX and run on live sensor streams inside the product, with latency monitored against the budget.
2020–23 Teaching an underwater robot where to look Bayesian models and a planner that ran onboard an autonomous underwater vehicle in Trondheimsfjord, steering it along a river-plume front instead of surveying evenly.
2022–25 Planning a long way ahead, around obstacles An RRT*-based planner that looks beyond the next waypoint: it trades information against distance over a cost valley while respecting obstacles and a time budget.
2021–23 Classifying water masses in three dimensions Adaptive sampling in 3-D: a vehicle porpoises through depth layers to classify river and ocean water, choosing each move by how much it reduces classification error.
2026 Machine learning, from first principles to production A curriculum of about fifty lessons written for colleagues, from autograd built by hand to GPT, LoRA, RAG and MLOps, alongside a Python course I wrote and taught.