There seems to be an endless number of new technologies to learn, and only limited time. To use my time most effectively, I’m planning to work on a single project that combines as many of the new techonologies that I can sensibly put into it. And instead of starting from scratch, I’m using as a basis a very nice codebase that comes with high quality data (link)
Dagster, Hopsworks, BentoML, K8s (and the usual parafelia: Helm etc.), Azure, Terraform, HuggingFace, Streamlit, Jax, Pytorch2(lightning), timescaledb, influxdb