Research Scientist (Causal Mathematics)

Location: London (hybrid/remote flexibility within the United Kingdom)

Type: Full-time

Team: Research

Focus: Causal discovery theory · statistical inference · algorithm design · causal data modeling

Why RootCause.ai

Most of the AI stack is correlation machinery. They can't answer "what if" questions. They hallucinate because they have no grounded model of reality. RootCause.ai is solving this problem from first principles. We have proven we can build causal inference systems that actually work. Now the question is: how far can this go? Can we ground abstract concepts through causal structure? Can we build systems that reason through interventions and counterfactuals? Can we create AI that knows what it knows versus what it's guessing? These are fundamental questions. We have the resources to pursue them seriously: venture backing, real customer data, production infrastructure, and a team that's already proven they can turn causal theory into systems that work.

The role

This is a research role focused on validating novel causal discovery methods and pushing theoretical boundaries. You will be working on fundamental questions in causal inference, representation learning for causal structure, bridging theory and real-world application. Validate core mechanisms rigorously. Prove what works and why. You will need to be comfortable with open-ended research problems without clear solutions. Time split: ~70% theory and derivation, ~30% Python prototyping and validation. Validated algorithms are handed to engineering for production implementation. This role is about getting the fundamentals correct, not production code.

Who you are

• Theoretically rigorous: You work through proofs, derive algorithms from first principles, reason about identifiability and convergence.
• Research taste: You know which problems matter. You can distinguish fundamental questions from academic busy-work. You see what's interesting about a problem before anyone tells you.
• Independently driven: You don't need specifications. Give you a research direction and you figure out what to test, how to validate it, what's missing. You expand ideas, often improving the original concept. You're the person who reads something and immediately thinks "but what about..."
• Exceptionally fast: You prototype quickly - days to test an idea, not weeks. You know what experiments actually prove something vs. what's just motion.
• Startup mentality: You care about research that matters. If something works, you want it solving real problems, not sitting in a drawer. Academic pace frustrates you. You tolerate ambiguity and move forward anyway.

What you’ll do

• Theory & algorithm development: Develop novel causal discovery algorithms grounded in statistical theory. Work through mathematical derivations: identifiability conditions, convergence guarantees, sample complexity bounds, asymptotic properties. Reason about computational complexity and algorithmic trade-offs (exactness vs. approximation, sample efficiency vs. runtime). Design causal data modeling components: conditional probability distributions, structural equations, intervention operators, counterfactual primitives.
• Prototyping & validation: Implement algorithms in your language of choice with clear, reproducible code. Build evaluation harnesses: synthetic data generators with known ground truth, benchmark comparisons, stress tests. Validate on real RootCause customer datasets (manufacturing, logistics, telecommunications). Document assumptions, limitations, failure modes, and theoretical guarantees.
• Research collaboration: Work independently on fundamental problems you identify. Collaborate closely with founding team on specific algorithmic challenges RootCause faces. Translate customer problems into research questions and vice versa.
• Research output: Publish findings (conferences/journals encouraged). Write detailed internal technical documentation. Present work to team and contribute to research roadmap.

Must-have

• Strong mathematical foundation: real analysis, probability theory, statistics (estimation, hypothesis testing, asymptotic theory, model selection).
• Algorithm design: ability to derive algorithms from theoretical principles and reason about their properties.
• Algorithm prototyping: implement algorithms from scratch, build experiments, validate results.
• Research communication: write clear technical documentation and research papers.
• Intellectual curiosity: ability to identify important problems and work through them independently.

Strongly preferred

• PhD in mathematics, statistics, or closely related field (or equivalent research experience).
• Familiarity with causal graphs, conditional independence, identifiability, interventions, structural causal models.
• Experience implementing algorithms in your programming language of choice.
• Publications in causal inference, statistics, or machine learning.
• Experience deriving algorithms: from theory → implementation → validation pipeline.
• Understanding of computational complexity: Big-O analysis, parallelization potential, scalability considerations.

Good to have

• Deep knowledge of specific causal discovery families (PC, GES, NOTEARS, LiNGAM, etc.).
• Probabilistic modeling (graphical models, Bayesian methods, variational inference).
• Numerical optimization (gradient methods, convex/non-convex optimization, regularization).
• C++ or Rust (helpful for understanding production constraints during handoff).
• Experience with large-scale data (Arrow/Parquet, columnar formats).
• Open-source contributions or strong technical writing.