Karthik Ranganathan is cofounder and co-CEO of Yugabyte. He works on distributed databases and agent data infrastructure, including the company’s Meko platform for sharing context and knowledge across agents.[1][2]
Start with Agent Memory Is Solved. Agent Learning Isn’t. →
Career
Ranganathan founded Yugabyte with Kannan Muthukkaruppan and Mikhail Bautin after work at Facebook. The team began building YugabyteDB in 2016. He is now co-CEO. His own 2019 presentation identifies earlier work at Nutanix, Facebook and Microsoft and study at IIT Madras and the University of Texas at Austin.[1][5]
Contributions
Bring distributed operation to a SQL database
His Yugabyte launch article introduces the company’s public beta and explains the team’s goal of handling business-critical data in cloud-native applications. The accompanying distributed-PostgreSQL presentation describes a design that combines PostgreSQL’s query layer with distributed storage, replication and transaction mechanisms. The work is a company and engineering-team effort; Ranganathan presents its architecture as a founder.[3][5]
Explain where shared knowledge belongs
In a company Q&A with Downing, Ranganathan discusses datapacks as portable, project-scoped stores rather than a single general database. The examples concern giving a team a consistent account of prior decisions and customer context. This makes retrieval and controlled sharing part of the application’s infrastructure, alongside the models that use it.[4]
Ideas
Ranganathan argues that passing a final document between agents loses failed attempts, constraints and reasoning. Meko’s datapacks store scoped context and shared knowledge. Agents can keep private working information and promote useful material for others. Heather Downing demonstrates the workflow in their joint session.[2]
Tools, writing & talks
Writing, research & talks
Start here
Ranganathan’s argument for shared context, followed by Downing’s Meko demonstration.
Start with handoff examples to understand what agents lose between sessions. The demonstration explains the proposed workflow without requiring installation.
Co-presenter with Heather Downing[2]Yugabyte's co-CEO argues shared, governed, promotable context is the missing layer for agent teams, and reports tuning a RAG pipeline one variable at a time to lift faithfulness from 14% to 82% with far less context, plus a demo of cross-agent resume.
Catalog workOriginal source- Yugabyte Has Arrivedyugabyte.com
The founder’s public-beta announcement and account of the database’s goals.
Author[3] - Distributed PostgreSQL on Google Spanner Architecturepostgresconf.org
A technical presentation with Mihnea Iancu about YugabyteDB’s distributed architecture.
Co-presenter[5]