The rapid expansion of the electric vehicle (EV) market highlights the crit- ical need for accurate State-of-Health (SoH) estimation to ensure battery reliability, safety, and lifecycle-aware operation. However, acquiring large- scale, high-quality operational data remains a major challenge. In real-world driving scenarios, on-field data are scarce, sampled at varying frequencies, and highly heterogeneous across different vehicle models. Furthermore, daily usage is dominated by partial charge and discharge segments, rendering clas- sical full-cycle health indicators impractical. This thesis presents a robust, data-driven Transfer Learning framework for estimating EV battery SoH using only pack-level signals (voltage, current, temperature, and State-of- Charge) available during normal driving operations.