If you can develop plasma control algorithms that are robust, constraint-aware, and ready for experimental operation, covering scenario design, supervisory control, and closed-loop regulation of core profiles, edge and divertor regimes, and off-normal response, we would like to hear from you. You will also contribute to subsystem digital twins and data pipelines that enable model development, validation, and deployment.
Key Responsibility Areas:
• Develop, test, and iterate plasma control algorithms for Eos PCS, including multi-rate feedback, supervisory logic, and constraint handling.
• Build AI/ML assisted control modules, such as learning-augmented MPC, reinforcement learning in simulation, physics-based models augmented with data-driven corrections, and adaptive control and auto-tuning across operating conditions
• Develop and deploy real-time state estimation and multi-diagnostic data fusion to produce control-grade plasma state, including robust profile reconstruction and regime indicators suitable for closed-loop operation.
• Develop anomaly detection and early-warning indicators across plasma and plant signals, with clear thresholds and graded mitigation actions.
• Partner with plasma physicists to translate physics goals into control objectives, observables, constraints, and validation tests.
- Digital twins and modeling:
• Design and maintain digital twin models for plasma control loops and coupled plant subsystems, enabling simulation, optimization, and controller development and validation.
• Build model validation workflows, including post-shot reconstruction, scenario sweeps, sensitivity studies, and uncertainty-aware comparisons against experimental data
- Data and integration:
• Build and maintain data pipelines for time-series control and diagnostic data, including labeling, archiving, and training and inference workflows.
• Integrate algorithms into the control stack in collaboration with EECS, ensuring compatibility with real-time constraints and low-latency inference where needed.
• Maintain high software quality: version control, test coverage, reproducible environments, and operational readiness documentation.
Ideal Experience & Skillsets:
Required qualifications:
• BS or MS in Electrical Engineering, Computer Science, Physics, Applied Mathematics, or related field.
• 3 or more years of experience developing control algorithms, ML models, or real-time scientific software for complex physical systems.
• Strong programming skills in Python, MATLAB and C++.
• Experience with time-series data processing, signal conditioning, system identification, and validation on real data.
• Practical familiarity with modern ML tooling (PyTorch, TensorFlow, or JAX) and deploying models into production code paths.
• Working knowledge of modern control methods, such as state estimation, constrained optimization, MPC, and robust control.
Preferred qualifications:
• Experience with plasma control, stellarators or tokamaks, or adjacent fields such as accelerators and large scientific infrastructure.
• Prior work on digital twins, real-time simulators, or physics-informed ML for physical systems.
• Experience integrating with experimental control and data systems (for example EPICS, MDSplus, SCADA, OPC UA).
• Familiarity with low-latency deployment and inference engines such as ONNX Runtime and TensorRT.
• Background in any of the following plant subsystems is a plus: power electronics, cryogenics, vacuum and pumping, high-power RF systems, and safety controls.
Tools and platforms:
• Languages and frameworks: Python, MATLAB, C++, PyTorch or TensorFlow or JAX, Jupyter
• Data infrastructure: MDSplus, EPICS, HDF5, InfluxDB, OPC UA, Kafka or ZeroMQ
• Dev workflows: Git, Docker, MLflow, CI pipelines
• Simulation and modeling: MATLAB Simulink, Modelica, FMI, and custom physics solvers
Company Benefits:
- Salary range $120,000-$160,000
- Comprehensive health benefits (e.g. medical/dental/vision)
- Employee equity stock options
- 20 days PTO
Requirements:
- Ability to occasionally lift up to 50 lbs.
- Ability to perform activities such as typing, standing, or sitting for extended periods of time.
- Willingness to occasionally travel or work required nights/weekends/on-call.
- Ability to work in a facility that contains industrial hazards including heat, cold, noise, fumes, strong magnets, high voltage, high current, pressure systems, and cryogenics.
Skills Required
- BS or MS in Electrical Engineering, Computer Science, Physics, Applied Mathematics, or related field.
- 3 or more years of experience developing control algorithms, ML models, or real-time scientific software for complex physical systems.
- Strong programming skills in Python, MATLAB and C++.
- Experience with time-series data processing, signal conditioning, system identification, and validation on real data.
- Practical familiarity with modern ML tooling (PyTorch, TensorFlow, or JAX) and deploying models into production code paths.
- Working knowledge of modern control methods such as state estimation, constrained optimization, MPC, and robust control.
- Ability to occasionally lift up to 50 lbs.
- Willingness to occasionally travel or work required nights/weekends/on-call.
- Ability to work in a facility that contains industrial hazards including heat, cold, noise, fumes, strong magnets, high voltage, high current, pressure systems, and cryogenics.
- Experience with plasma control, stellarators or tokamaks, or adjacent fields such as accelerators and large scientific infrastructure.
- Prior work on digital twins, real-time simulators, or physics-informed ML for physical systems.
- Experience integrating with experimental control and data systems (for example EPICS, MDSplus, SCADA, OPC UA).
- Familiarity with low-latency deployment and inference engines such as ONNX Runtime and TensorRT.
- Background in plant subsystems (power electronics, cryogenics, vacuum and pumping, high-power RF systems, safety controls).
What We Do
Thea Energy is a fusion energy company dedicated to creating a limitless source of zero-emission energy for a sustainable future. By reinventing the stellarator using arrays of mass-manufacturable magnets and dynamic software controls, they aim to commercialize scalable and economical fusion power systems, transitioning from scientific research to practical, commercial operations.
