Model choice becomes harder when a workflow can succeed in more than one way.
Cost, latency, and output quality rarely point to the same default for every request.
Who you’d be doing this for
“We keep adding exceptions because no single model is reliably the right call.”
Valentina Ribeiro · Principal AI Engineer
Maintains several production LLM workflows and currently encodes model fallback rules in application code.
What is at stake
Model switching rose 62%, but thin telemetry and customer feedback disagree on the reason. You have to weigh a narrow, measurable experiment against the temptation to solve every model-choice problem at once.
Why it isn’t already fixed
Every obvious fix costs something else. That’s the part you’d have to decide.
- broad flexibility vs. testable customer value
- offline quality vs. workflow completion
- fast platform reuse vs. evidence-led direction
Why Databricks
Foundation Model APIs expose partner and self-hosted models through a unified serving layer for enterprise AI workloads.
Written with these in mind
Not your kind of problem? 34 more at Databricks, or browse every organization.
This is the setup. The work is inside.
Running it puts you in the room: the full situation and its constraints, stakeholders who push back in their own words, and the decisions that are yours to make. What you produce becomes a Day One Plan — work you can show someone instead of describing.