[Remote] Principal AI Engineer, Closed-Loop Control
Note: The job is a remote job and is open to candidates in USA. 4MP Inc is building the intelligence layer for precision manufacturing. The Principal AI Engineer will design and build AI systems for closed-loop control in manufacturing, working closely with the founding team to develop practical solutions that integrate real-world sensor data and machine behavior.
Responsibilities
- Design and build AI systems that connect sensor data, machine behavior, manufacturing context, and closed-loop correction
- Work directly with the founding team on systems that interact with real machines, real sensors, and real manufacturing data
- Sensor intelligence — calibration, fusion, uncertainty quantification, and drift detection
- Physical AI modeling — machine behavior, error modeling, and physics-informed learning
- Manufacturing context — interpreting CNC programs, machining intent, and process state
- Closed-loop correction — learning systems that improve from real correction-to-outcome feedback
- Scaling intelligence — transferring learned knowledge across machines and deployments
Skills
- 6+ years in AI/ML with substantial work in control, optimization, reinforcement/imitation learning, or inverse problems
- Experience turning measured error or state into corrective action — closed-loop systems where model outputs change physical behavior
- Strong optimization and control foundation (numerical optimization, MPC, or learned control)
- Comfort working against real, noisy feedback — drift, delay, partial observability, and safety constraints
- Strong Python and production-grade ML engineering (PyTorch)
- Mathematical maturity in optimization, control theory, dynamics, and probability
- Degree in CS, Robotics, EE/ME, Physics, or Applied Math — or equivalent demonstrated work
- Reinforcement / imitation learning for control ; differentiable simulation
- Model-predictive control , trajectory optimization, system identification
- Differentiable physics / rendering ; physics-informed learning (PINNs)
- Uncertainty-aware decision-making — Bayesian methods, conformal prediction, risk-aware control
- Working with geometric / 3D representations as input to control (bridges to the perception side)
- CNC / machining physics : tool deflection, thermal error, material removal, fixturing
- CAM, G-code, toolpath generation
- M.S. or Ph.D. in control, robotics, optimization, or related
Company Overview