In this talk, I will provide an overview of my past, ongoing, and future ML for systems research, starting with my postdoc years at MIT, where I worked on benchmarking learned indexes, learning cardinality estimates, and developing SageDB, an instance-optimized Data Analytics System. I will also talk about my experience at Amazon, discussing my contributions to intelligent scaling in Amazon Redshift and Text2SQL translations. I will shine some light on the characteristics of cloud data warehouse workloads, and will emphasize the potential of (sub) query caching. Finally, I will talk about my ongoing and future research at UTN, where we work on simplifying data loading with LLMs, correlation-aware column compression, and optimizing queries by learning from query history.