Field notes from the data terrain
Charting the ground between raw data and working AI systems.
Let's work together on your AI journey
Now: Databricks Genie AI MCP adoption Use Cases
Note of the week
Most model regressions I've debugged trace back to upstream schema drift, not the model itself—the pipeline feeding it changed shape quietly, and nobody noticed until the outputs did. Treat your data contracts with the same rigor as your APIs, because they break just as often and are usually caught much later.
What I do
Data pipelines
Ingestion, transformation, and orchestration that hold up under real production load.
AI integration
Wiring Claude and other models into products as features, not demos.
Applied AI features
The interface layer on top — the part your users and customers actually touch.