Worked on enhancing machine learning model deployment within the XENONnT/straxen repository by upgrading the TensorFlow and Keras model loading protocol. The approach involved adding support for Keras 3, removing the deprecated tf:// protocol, and registering classes directly from keras.tar.gz archives to streamline packaging and reduce load-time errors. Addressed compatibility issues with NumPy 2 and refined the posrec plugin to ensure stable integration of updated models. Leveraged Python development skills and configuration management expertise to broaden compatibility, reduce maintenance overhead, and enable faster adoption of modern machine learning models in the straxen data processing workflows.