Battery Machine Learning Algorithm Engineer

Summary

Envision what you could do here. At Apple, we believe new ideas have a way of becoming extraordinary products, services, and customer experiences very quickly. Bring passion and dedication to your job and there's no telling what you could accomplish.

Minimum Qualifications

Experience in machine learning (RNNs, LSTMs, Transformers), control theory and optimization (MPC, Kalman filters, convex/nonconvex optimization) Experience in MATLAB/Simulink Experience in C/C++ for embedded systems and Python & ML frameworks (PyTorch, TensorFlow)

Preferred Qualifications

PhD and 10+ years relevant industry experience Mathematical modeling experience for nonlinear dynamical systems Experience with physics-informed neural networks or hybrid physics/ML approaches Experience with MPC implementation and real-time optimization

Description

We are looking for an engineer who will develop novel machine learning algorithms, control strategies, and on-device implementations to improve the utility and intelligence of battery systems.

Responsibilities

Develop ML and adaptation algorithms for State of Health, State of Power, and State of Charge (SOH/SOP/SOC) estimation — characterizing degradation mechanisms (capacity fade, impedance growth) and peak power capability, with robustness across temperature, cell aging, dynamic loads, and inaccurate current sensing. Build ML-enhanced predictive models of user load patterns, system power demand, and coupled electro-thermal dynamics, enabling proactive, thermal-aware battery algorithms that extend runtime and improve power efficiency. Design advanced control algorithms (MPC, adaptive, learning-based) and multi-objective optimization frameworks to optimize battery management and power delivery in real time — balancing runtime, cycle life, thermal and system constraints, and safety margins. Collaborate with system teams to co-optimize battery algorithms with device power management and thermal systems. Develop physics-informed ML approaches that combine physics-based models with data-driven methods to improve generalization and ensure physically plausible predictions. Deploy algorithms on resource-constrained embedded platforms using model compression (quantization, pruning, distillation) and fixed-point arithmetic, with graceful fallback under sensor noise, missing data, and distribution shift.