Rohit Ghumare

Ghumare’s current About page identifies work with the team building iii, a distributed backend organized around workers, functions, and triggers. His writing and speaking cover agent harnesses, Kubernetes, inference systems, security, and observability. His portfolio also lists open-source courses and community teaching.[1][4]

AgentMemory captures coding sessions and makes decisions and fixes searchable later. A shared store lets different supported coding tools retrieve the history instead of rebuilding it from a single application’s conversation. Default keyless retrieval uses keyword search; optional local or remote embeddings add semantic matching. A local viewer makes saved material inspectable. The distinction matters because a query that depends on meaning rather than matching terms may behave differently before embeddings are configured.[3]

Ghumare’s portfolio links AI Engineering from Scratch as a course built in code, alongside tools for managing agent skills and connecting Kubernetes to MCP clients. The combination gives readers an implementation-focused route into agent engineering: study the components, then inspect tools that use them. AgentMemory similarly exposes capture and recall as explicit operations, so persistence can be checked rather than assumed from a successful chat response.[4][3]