Physical AI Engineer (Control, Optimization, Validation)

PassiveLogic•Holladay, UT

47d•Hybrid

About The Position

PassiveLogic is seeking a Physical AI Engineer to develop, optimize, and validate autonomous control systems, AI models, and levels of autonomy for their fully autonomous platform for buildings. This role involves combining control theory, optimization, machine learning, physics-based simulation, and systems validation. The engineer will work on digital twins, autonomous reasoning, building physics, equipment behavior, and human comfort, developing production-quality algorithms tested in simulation, laboratory systems, and real buildings. The goal is to enable buildings to understand their physical state, learn from experience, and autonomously optimize their operation, contributing to reduced carbon footprints and optimized building performance.

Requirements

MS or PhD in control engineering, computer science, robotics, applied mathematics, mechanical engineering, or a related field.
Demonstrated expertise in AI development, scientific machine learning, optimal control, reinforcement learning, multi-agent systems, physics-informed machine learning, and optimization theory.
Strong technical background in control theory, model predictive control, state estimation, and system identification.
Strong programming skills in Python, C++, Swift, or a similar language.
Strong analytical, debugging, root-cause analysis, communication, and collaboration skills.

Nice To Haves

Experience with automatic differentiation and differentiable programming.
Experience with software design, design patterns, and software architecture.
Knowledge of building science, HVAC systems, thermodynamics, energy modeling, or grid-interactive controls.
Experience with software-in-the-loop and hardware-in-the-loop testing.
Experience with formal methods, probabilistic modeling, and graph neural network.
Experience with building energy-modeling tools such as Dymola and EnergyPlus.
Experience in vector, SIMD, and tensor computational methods.

Responsibilities

Design and implement autonomous control strategies that use physics-based digital twins for prediction, optimization, and decision-making.
Translate comfort, energy, equipment life, and operational requirements into control objectives, constraints, and cost functions.
Develop scalable optimization algorithms using methods such as stochastic gradient descent, coordinate descent, distributed optimization, Bayesian methods, and evolutionary algorithms.
Design and implement physics-based automated fault-detection and diagnosis algorithms.
Develop fault-tolerant control methods that support degraded operation and system recovery.
Integrate control and learning algorithms into production systems while meeting reliability and real-time performance requirements.
Develop physics-informed predictive and learning models using deep learning, reinforcement learning, and transfer learning.
Develop autonomous agents, state-estimation methods, model adaptation, and control-correction algorithms.
Create learning methods that remain reliable under noise, outliers, sparse data, model uncertainty, and changing system behavior.
Define performance metrics, acceptance criteria, and automated tests for autonomous capabilities and levels of autonomy.
Test system behavior under disturbances, incomplete observations, model divergence, and equipment, sensor, or communication failures.
Verify fault-detection accuracy and validate fault-tolerant control, degraded operation, and recovery behavior.
Compare simulation results with analytical solutions, laboratory measurements, and real-building data.