Anna Rohova

Developed and productionized EFC-Axis Universal, a source-agnostic battery-wear metric unifying heterogeneous OEM fleet and laboratory data, selected as the production model after benchmarking multiple architectures. Designed physics-informed ML models for state-of-health and remaining-useful-life estimation, including Bayesian PINN ensembles, teacher–student distillation, hierarchical Neural ODEs, Clockwork-LSTM, and continual learning with EWC. Built physics-based battery simulations in PyBaMM (DFN), modeling SEI growth, particle cracking, and lithium plating; used simulations as synthetic ground truth to validate data-driven models. Engineered end-to-end pipelines for real EV telemetry (including a 300-vehicle fleet and a large OEM battery-pack dataset), created a defect taxonomy for production BMS data, and identified silent data defects (e.g., phantom-cycle bug and duplicate-vehicle data leak) that materially affected evaluation.

Supported university-level mathematics education through problem-solving sessions and student guidance, focusing on Differential Equations, Linear Algebra and Mathematical Analysis, and Mathematical Analysis 1. Helped communicate complex topics clearly while adapting explanations to different levels of understanding.

Designed individualized learning approaches, course methodology, and teaching strategies based on student performance analysis. Taught advanced mathematics and problem-solving across different levels for multiple learners.