Deploying a fleet of nuclear power plants is one of the most ambitious and consequential undertakings in the global energy transition — and The Nuclear Company is doing it. You will join a small but world-class Applied Research and AI team and work on genuinely hard, open research problems at the intersection of AI and large-scale infrastructure: how do you optimize construction across a fleet of simultaneous sites, allocate capital intelligently under deep uncertainty, and keep a distributed critical infrastructure secure? These are not incremental problems — they sit at the frontier of applied AI research, with real operational stakes and the potential to reshape how the energy industry is built.
Research & Modeling
Problem Formulation: Translate complex operational processes into well-defined research problems; identify the right modeling approach for each domain and build the case for why it will work in practice.
Simulation & Evaluation: Build simulation environments that faithfully represent our operational processes — construction scheduling, portfolio sequencing, security operations — and can be used to train, evaluate, and iterate on decision-making models.
Empirical Research: Design rigorous experiments, maintain reproducible codebases, and communicate results clearly in internal reports and, where the research warrants it, external publications.
Some of the exciting topics you are likely to work on include:
Construction Schedule Optimization
Schedule Optimization: Develop models that optimize construction scheduling across multiple concurrent sites — minimizing schedule variance, resource idle time, and cascading delays across a growing fleet of projects.
Dynamic Rescheduling: Design approaches that adapt scheduling decisions in real time to disruptions — supply chain delays, labor fluctuations, permitting hold-ups — learning from historical project data to improve over time.
Site Portfolio Optimization
Portfolio Decision Systems: Build models that inform how we sequence site development and allocate capital across a growing fleet — accounting for regulatory milestones, capital constraints, and correlated risks across sites.
Uncertainty Quantification: Develop approaches that account for uncertainty in key inputs — permitting timelines, cost distributions, grid demand forecasts — to produce portfolio decisions with bounded downside.
Offline / Batch RL: IQL, CQL, TD3+BC, Decision Transformer, or similar methods — directly relevant given limited online interaction in our deployment environments.
Combinatorial Optimization + ML: Graph neural networks for scheduling or routing (GCN, attention-based), neural combinatorial optimization, or hybrid learned/exact solver approaches.
Multi-Agent RL: MADDPG, QMIX, MAPPO, or related methods for multi-site coordination and adversarial security formulations.
Stochastic / Robust Optimization: CVaR-constrained RL, distributionally robust MDPs, or chance-constrained programming for decision-making under uncertainty.
Production RL Deployment: Experience monitoring and retraining RL systems post-deployment, including distribution shift detection and safe policy update procedures.
Domain Exposure: Construction project management, infrastructure operations, energy industry, electricity markets, nuclear power, industrial control systems, or physical/cyber security for critical infrastructure.
Game Theory: Familiarity with game-theoretic frameworks — strategic interactions, mechanism design, adversarial dynamics, or equilibrium concepts — provides useful modeling perspective for problems where multiple decision-makers interact, compete, or coordinate.
Behavioral Science: An understanding of how people actually make decisions — including cognitive biases, bounded rationality, and responses to uncertainty — is valuable when building systems that work alongside human decision-makers.
