Chinmay Relkar is a software engineer at Clarisights. He builds developer tools and writes about agent harnesses and the organizational context that helps AI systems make useful decisions.[1]
Start with Organizational Cognition: Why the Next AI Moat Won't Be Intelligence →
Education
BE, Information Technology
2015 – 2019
[1]Links
WebsiteSoftware Engineer · Clarisights
Career
Relkar’s work ranges from mobile and internal infrastructure to production AI systems. At Unacademy and Cohesive he built generative-AI infrastructure and products; his current Clarisights work includes AI and orchestration. His earlier roles include Amazon, 100ms, Schlumberger and internships in mobile development and data science.[1]
Contributions
His agent-harness presentation follows a progression from a prompt loop to tool contracts, executable interfaces, context management, isolation and evaluation. Each layer addresses a concrete failure of the previous system. The presentation treats a model and its surrounding execution environment together when explaining what an agent can reliably accomplish.[4]
Ideas
Organizational Cognition argues that AI products increasingly compete over a company’s decision patterns: which reviewer to involve, how to resolve conflicting sources and when an exception is justified. The essay separates memory, reasoning, coordination, learning, delegation and governance. Relkar labels observed product decisions separately from interpretations and forecasts. The framework is an argument about product direction, not experimental proof that stored organizational context guarantees better decisions.[3]
Tools, writing & talks
Writing, research & talks
Start here
Organizational Cognition: Why the Next AI Moat Won't Be Intelligencechinmayrelkar.github.ioAn essay distinguishing organization-specific judgment from facts, workflow definitions and generic model intelligence.
Start with the definition and export test, then compare the taxonomy with one workflow you know. Familiarity with enterprise AI helps; read the stated observations and forecasts as different kinds of claims.
Author[3]Original source- The Evolution of the Agent Harnesschinmayrelkar.github.io
A chronological explanation of the layers around a coding-agent model.
Author[4]