Research Engineer – Forecast Coherence & World Models Full Time

Description
Metaculus is the premier public forecasting platform with over 2,800,000 forecasts on over 20,000 questions. Metaculus is “epistemic infrastructure” like Wikipedia and Our World In Data, providing the world with trusted, accurate information on a variety of questions, particularly in science and technology. Our mission is to enable everyone to model, understand, predict, and navigate the world’s most important and complex challenges.

As a Research Engineer focused on Forecast Coherence & World Models, you’ll pioneer the development of technologies that make forecasts and their underlying world models transparent and accessible. You’ll work at the intersection of software engineering and epistemics, building tools that help domain experts and forecasters create more coherent predictions, and support better decision-making on complex, cross-disciplinary problems.

This role emphasizes independent experimentation and rapid prototyping. You will run LLM experiments, design and test new approaches, and build MVPs to validate ideas. You will also design methods to elicit, aggregate, and display the relationships between forecasts (e.g. causal connections, necessary conditions) while leveraging AI to assist in world model construction and identify missing variables that could improve forecast quality.

Responsibilities
Independently scope and run experiments with LLMs, prompt engineering, and novel approaches to eliciting and aggregating forecast relationships
Build MVPs and prototypes to test and iterate on ideas quickly
Design and implement methods for eliciting, aggregating and visualising coherence rules across forecasts, including causal/correlational relationships, martingale processes, necessary/sufficient conditions, etc.
Develop intuitive interfaces that make complex forecast relationships and world models accessible to stakeholders
Build AI-powered tools for world model elicitation and legibilization, including autonomous coherence rule detection and latent variable suggestion systems
Integrate new capabilities into the main Metaculus platform and potentially build stand-alone tools for specialised use cases
Our Tech Stack
Frontend: JavaScript, TypeScript, React, Next.js, D3, Tailwind CSS
Backend: Python, Django, PostgreSQL, Redis, Sentry, Heroku, AWS
Collaboration: GitHub, Slack, Notion, Figma

Requirements
Clean Code Practices: Commitment to writing maintainable, testable, and well-documented code
Problem Solving: Strong analytical skills with the ability to tackle ambiguous problems in novel domains
AI integration: prompting, RAG, fine tuning, etc.
Communication: Excellent written and verbal communication skills, essential for remote collaboration
Mission Alignment: Genuine interest in improving humanity’s ability to predict and navigate complex challenges