Senior Software Engineer, Simulation & Forecasting

The role

You turn AI-generated research into decisions people can act on: the report they read, and the world model behind it. Here, a world model is an executable simulation of the decision, with its causes, effects and uncertainties. It tests scenarios and forecasts what will happen.

Why this role

• You build the part of the product that turns evidence into a decision.
• Your forecasts are scored against real outcomes, so their quality is measured, not claimed.
• Simulation, probabilistic forecasting and LLMs in one role, from the model to the report.

What you will do in your first year

• Build the report layer: report formats, charts of the evidence, digitally signed output, and an assistant that answers questions about a finished report.
• Build the engine that creates world models from research results, with parameter ranges, cause-and-effect links and scenarios, validated against reference data.
• Add calibrated forecasts: questions registered in advance, probability forecasts with measured quality (for example Brier scores), and comparison with what actually happened.

What you bring

• Several years designing and owning production software in Python. Show us one.
• Quantitative modelling: simulation (Monte Carlo, system dynamics or agent-based), probabilistic forecasting, uncertainty.
• Forecast evaluation: proper scoring rules, calibration, backtesting.
• Turning model output into clear reports and charts.
• Comfort with LLM pipelines and with directing AI coding agents: you turn a vague goal into a precise work order an agent can build from, and you can show us a time you caught an agent’s code or tests being wrong.

Nice to have

Bayesian modelling (PyMC, Stan); decision analysis; forecasting tournaments or platforms; document generation (PDF, HTML); domain knowledge in energy, industry or finance.