SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Treefera is a first-mile intelligence platform delivering granular visibility into global agricultural and soft commodity supply chains. The company uses AI, satellite data, and quantitative modelling to help organizations understand risk, cost, performance, and exposure at the point of origin.
In this role, you will own the models and data pipelines that transform weather and satellite data into the risk and market signals customers rely on. You'll take ambiguous questions—such as where agricultural stress is building this season and how confident we can be about it—and build end-to-end data pipelines from ingestion through modelling and evaluation to customer delivery.
Key responsibilities include:
- Building and shipping forecasting models for environmental and risk signals, from agricultural stress indicators to weather and climate volatility, and taking responsibility for live performance.
- Extending the weather platform by adding new forecast products and capabilities to an established staged pipeline (ingestion, standardisation, spatial aggregation, climatology, indices, stress scoring).
- Working with satellite data across optical and radar missions to build vegetation stress signals, landcover classifications, and land-surface conditions.
- Taking research from prototype to production: building infrastructure, designing failure modes, and turning one-off work into orchestrated, reproducible data deliveries.
- Shaping the AI team's modelling practices by improving experiment design, evaluation protocols, documentation, and uncertainty treatment; communicating methods and limits clearly to technical and commercial colleagues.
You will report to Tommy Lees and work within the AI pod, partnering closely with the Science team on methods, Engineering on platform infrastructure, and Product and Market Intelligence on customer signal requirements.
Success milestones: by 30 days, you'll have local pipelines running and understand system weaknesses; by 60 days, you'll have delivered your first improvement (new index, better evaluation, or forecast product); by 90 days, that work runs in production with external reliance; by six months, you own a signal domain end-to-end; within a year, you've shipped materially better forecasting capability with evaluation evidence.
Requirements:
- Deep experience building deep learning and statistical models for time series or spatial data, with real projects you can walk through in detail, including failures.
- Fluency in the Python scientific stack (PyTorch, scikit-learn, scipy, xarray) and reproducible practices (version control, experiment tracking, orchestration, cloud infrastructure).
- Ability to interrogate data before modelling, state assumptions clearly, and communicate uncertainty straightforwardly to stakeholders who will act on results.
- Demonstrated end-to-end problem ownership: taking ambiguous questions to validated models to production pipelines.
- Ownership and understanding of all code shipped, including AI-assisted development.
Desirable:
- Experience with weather and climate data (reanalysis products, numerical weather forecasts, weather station records, forecast verification).
- Experience with remote sensing datasets.
- Exposure to risk modelling, financial time series, commodity markets, backtesting systematic strategies, and understanding how models generate tradable signals.
Right to work: You'll typically need the right to work in the UK, as visa sponsorship is not generally available. If your situation differs, the company will discuss it.
