About the role
This is a senior, hands-on coding role.
You will lead the data and evaluation engineering for the benchmark and extend our proprietary platform, as part of a collaborative team that includes data scientists at NPL, AI partners at Applied Form, and materials scientists at Biophilica.
Your work will include, but won't be limited to:
- Mature the data pipeline that ingests, cleans and standardises literature, patent and in-house experimental data into a benchmark dataset of 5,000+ datapoints
- Implement the benchmark schema and Functional Materials Ontology, including annotation tooling, confidence scoring and contradiction tracking
- Build baselines and the evaluation harness with NPL: standard models (XGBoost, SchNet, CGCNN and others), reproducible train/test splits, uncertainty calibration and a benchmark API
- Ship the open release: a documented public dataset slice and harness on GitHub that third parties can run
What we're looking for
Essential : 5+ years of professional software engineering, with at least 2 years building ML systems in production or research
Expert-level Python (pandas, NumPy, scikit-learn, PyTorch) and strong software practice: testing, version control, CI, clear documentation
Experience building data pipelines from messy, heterogeneous sources, including PDF and text extraction
Solid grounding in model evaluation: train/test design, error metrics (MAE, R²), and uncertainty or calibration methods
Cloud experience, ideally Google Cloud Platform Azure, and/or Cloudflare
Able to work independently on a short, milestone-driven contract and deliver from day one
Desirable: Experience with scientific or materials data (Materials Project, MatBench, JARVIS) or graph neural networks for chemistry
Knowledge graphs, ontologies or neuro-symbolic approaches
LLM-assisted information extraction from scientific literature
Experience publishing open-source datasets, benchmarks or tools
Startup experience and comfort working across companies and disciplines
What we offer
A fixed-term contract of up to 8 months, January to August 2027, on a newly awarded Innovate UK project
Competitive salary
Semi-remote working, with regular collaboration days in London
Work alongside NPL, the UK's national measurement institute, on benchmark design and validation
Public credit on an open benchmark that researchers and AI developers will use
Occasional travel within the UK
A small, collaborative team making a real environmental impact