I build things to understand them.
Mostly diffusion models, graph neural networks and molecules, and lately the odd tool I wished existed.
A little about who I am

I'm Søren, an AI engineer doing my master's at DTU, drawn to where deep learning meets structure: diffusion models, graph neural networks and molecules.
Alongside my studies I worked at Ørsted, helping bring generative AI into the company and teaching much of its upskilling programme. Today I also co-run Botline, a small AI consultancy.
Most of what I build, I build to understand something properly. A few of those things are below.
- At Ørsted
- 4 yrs
- Live AI teaching
- 20+ h
- DTU, in progress
- MSc
Focus areas
- Diffusion models
- Graph neural networks
- Molecular ML
- Generative AI
Languages
- Python
- TypeScript
ML tooling
- PyTorch
Other
- Teaching & workshops
- AI strategy
Things I've built and taken apart

Finding gas in seismic data
Only 25 examples of the thing I had to find. Treating every column of every image as its own example turned them into more than 10,000, and small models did the rest.

Equivariant diffusion for 3D molecules
My bachelor's thesis: a diffusion model that generates 3D molecules, with PaiNN inside, built from scratch in PyTorch.

PolyPred
Predicting polymer properties from a single monomer SMILES string with an equivariant GNN on cheap, guessed 3D geometry. It matched a strong tabular baseline, and beat it on free volume.

Hours
A time tracker for my consultancy where logging hours is mostly confirming: narrow the week to what you worked on, and let git history and the calendar propose the entries.

CouchCode
Claude Code from the sofa, with a PS4 controller: buttons for keys, sticks for the pointer, my voice for the typing, and a rumble when Claude is done.
What I'm loving, and what I'm not
The professional charts, updated whenever my opinions are.
- 01No changeDeep understandingImplementing it until it makes sense
- 02No changeDiffusion modelsStill feels like magic, now with maths
- 03Up 2 since 2025Agentic codingFrom toy to daily driver
- 04Down 1 since 2025Molecules and GNNs
- 05Down 1 since 2025Teaching20+ hours on stage and still enjoying it
- 06Up 3 since 2025Legal and contractsSurprisingly satisfying
- 07New entryReverse engineeringThanks, Digimon
- 08No changeAdministrationFine. Necessary.
- 09Down 1 since 2025PowerPointPeaked in 2023. LLMs speak HTML now.
- 10No changeGovernance meetingsFun at first. Got old fast.
- 11No changeKaggle leaderboardsAsk me about PolyPred
- 12New entryTax registrationsThe tax hydra: answer one question, two more appear
- 13Down 3 since 2025SalesChasing people for money
- 14No changeLinkedInBottom of the charts, as always
Where I've been
Work
- 2026 to present
Co-founder @ Botline ApS
Co-founded a small AI consultancy with a friend. We build AI solutions for clients, and on the side built an AI intake platform for consultancies. With two of us, we cover everything ourselves, from architecture and development to contracts and administration.
- 2025 to 2026
Teaching assistant @ DTU
Introduction to Intelligent Systems, fall 2025. Helped students through exercises and projects. I also wrote and delivered a short lecture on DTU’s high-performance computing cluster, after finding that the existing introduction left a lot to be desired.
- 2021 to 2025
Student assistant @ Ørsted
Digital Strategy, 2023 to 2025. Part of a small team helping Ørsted find its footing with AI. I prototyped and evaluated use cases, worked through the security and governance questions that came with them, and built and ran a large part of the generative-AI upskilling programme, including more than 20 hours of live teaching. I also built a model for detecting subsurface gas in offshore seismic data.
R&D and Innovation, 2021 to 2023. Coordinated the team’s conference participation and research-proposal intake, facilitated engineering onboarding courses, and later reworked their course material.
Education
- 2026 to present
MSc, programme TBD @ DTU
Focusing on deep learning, especially diffusion models, graph neural networks and machine learning for molecules.
- 2022 to 2025
BSc, Artificial Intelligence and Data @ DTU
The foundation: machine learning, statistics, optimisation and software engineering, plus a few projects I took well beyond the brief.