From Recommendation to Action: Scaling Autonomous Databricks Optimization in the Enterprise
Thursday, August 6, 2026 | 11:00 AM PT / 2:00 PM ET | 30 minutes
Every Databricks optimization platform can tell you what should change. The harder question is: when should a system be trusted to make that change on its own?
This session follows the journey of an Unravel customer as we moved from surfacing Databricks platform optimization recommendations to safely applying them in production. Prajakta will talk about the engineering decisions, the guardrails, and the trust model that made autonomous optimization possible in an environment where every change carries operational risk.
- Why recommendations are only the beginning — the gap between identifying an optimization opportunity and actually making the change in production.
- What makes a recommendation actionable — the technical and operational checks that determine whether a recommendation is truly ready to execute.
- Building trust, one step at a time — moving from manual approvals to scoped autonomy instead of jumping straight to full automation.
- Auto-Apply in production — how recommendations are validated, executed, and continuously verified to ensure they deliver the expected outcome.
- Keeping humans in the loop — using a PR-based workflow so every automated decision remains transparent, reviewable, and auditable.
Platform engineering leaders, FinOps and DataOps practitioners, cloud architects, and engineering managers exploring how to introduce autonomous optimization into production systems without compromising reliability or governance.
A practical framework for moving beyond recommendations and building systems that can earn the right to act autonomously. We will share what worked, what did not, lessons learned, and the results from deploying automation in a large-scale enterprise Databricks environment where trust mattered just as much as savings.
Speaker
Prajakta Kalmegh
Prajakta has spent 21 years at the intersection of enterprise AI, cloud data infrastructure, and applied research. At Unravel, she leads the teams building the multi-agent systems that turn platform recommendations into validated, applied fixes.