I build software for bioinformatics, computational drug discovery and structural biology. Most recently an AI-driven annotation pipeline where I designed a voting system, and deterministic validators that overrule the AI. Open to AI / agentic engineering, data engineering, data science and scientific software roles in the Copenhagen area.
I'm a PhD researcher (Drug Design & Pharmacology, University of Copenhagen / SUND) and a GPCRdb developer. GPCRdb is the reference database for a receptor family that's the target of roughly a third of approved drugs, and I build the software behind it. Lately that includes an AI layer that has to earn a scientist's trust field by field, not just look good in a demo -- deterministic validators, audit trails, human-in-the-loop review.
GPCR Annotation Tools is an AI-driven annotation pipeline for GPCR structural data. Gemini proposes each field; a chain of deterministic validators can overrule any of them against an external database or a computed value, and every override is logged for human review. Started as a fork of the GPCRdb team's repository, all 214 commits are mine, 8 PRs merged upstream. Being deployed into GPCRdb; NAR manuscript submitted (first author). repository · merged PRs
The container platform GPCRdb runs on. Django + RDKit-PostgreSQL,
health-check-gated startup, multi-architecture images, and a
36-cell Python x Django x RDKit compatibility matrix that runs
automatically in CI on every push. Ten of my PRs are merged into
protwis/protwis (43 stars, 75 forks).
merged PRs ·
protwis_django_docker ·
postgres_rdkit_docker
A pre-computation pipeline for molecular modeling runs as a private codebase, with staged jobs on an HPC cluster under SLURM for the licensed Schrodinger Suite, explicit contracts between stages, idempotent and resumable, and about 250 automated tests plus a golden-fixture regression suite that catches drift after a vendor upgrade. No public repository to link.
LSTM models over 15.6 million viral genomes, from my MSc research -- sequence models in PyTorch, trained on HPC, tracking SARS-CoV-2 evolutionary dynamics across the full GISAID corpus. paper (DOI) · code snapshot (Gene, 2024, 916:148426, first author)
I co-developed Plasmer as third author, a Random Forest classifier for plasmid host-range prediction that other labs now run in production (Microbiology Spectrum, 56 citations, 1,100+ Docker pulls). repository
A 3,000-word Danish core-vocabulary Anki deck generated with IPA and audio.
Local Whisper transcription of Danish public-radio audio into subtitles.
A smaller Danish-learning helper tool.
A push-notification CLI for my own automation pipelines.
Auto-login for a campus network gateway, and where my coding started.
Core
AI / ML
Cloud (used, not deep)
If you're building AI that has to earn trust in a regulated or scientific setting, I'd like to hear about it -- reach out here on GitHub.
