Akhil Gudesa I build the controls that let AI agents do real work. At Databricks, I own the governance side of Unity AI Gateway: which models and tools agents may use, whose permissions apply, and how their actions are recorded. Outside work, I’m building a product for delegating work to AI agents that carry your context, learn from your corrections, and respect your boundaries. The question behind it: can an agent do the work the way you would, while keeping you in control? My six years at Databricks span query-engine performance, distributed data systems, and AI governance. I helped build Spark Connect’s JVM client from its first commit and its support for shipping user code to shared servers with session-level isolation. I’m the first-named inventor on US Patent 12,210,521 for short-query prioritization, co-presented at Data + AI Summit 2024, and represented IIT Indore at the ICPC World Finals in Moscow. From Bengaluru, now in Amsterdam. Usually cycling, snowboarding, or training for a triathlon when I’m away from the keyboard.
Since 2026, I own governance for Unity AI Gateway: policies covering models, tools, MCP servers and skills, plus permissions and audit trails. Previously worked on Spark Connect.
Worked on Spark Connect’s JVM client and support for shipping user code to shared servers, contributing to Apache Spark’s client-server architecture.
Worked on query-engine performance in C++ and Scala; part of the Databricks team behind the 2021 TPC-DS world record.
Research Assistant
Contributor — pgRouting / OSGeo
Engineering Intern
B.Tech, Computer Science and Engineering
