Nacho Martínez

Martínez works on agent memory, AI infrastructure and developer education at Oracle. His portfolio connects that work with open-source tools, workshops and the Oracle AI Developer Hub. His teaching connects database-backed persistence with runnable agent examples.[1][2][5]

Make memory an explicit part of an agent

With Richmond Alake, Martínez teaches how to build persistent agent memory using Oracle AI Database, LangChain and language-model pipelines. The course develops a memory manager, retrieves tools through semantic search, and extracts and consolidates information for later sessions. Its exercises distinguish storing a conversation from maintaining structured memory that a subsequent run can retrieve and update.[2]

Publish an environment people can inspect

The Oracle AI Developer Hub collects notebooks, reference architectures and runnable examples for Oracle AI Database and OCI. Martínez lists it among his work. His portfolio also distinguishes original tools from forks that replace an existing agent’s persistence layer with Oracle storage, making the integration work visible without presenting the underlying agents as his inventions.[1][6]

In Total Recall, Martínez treats the model as a fixed reasoning component and places memory, tools, retrieval and orchestration in the surrounding harness. Context cards organize intent, relevant facts and recent exchanges. Tools can be retrieved for the current step rather than all loaded together. He also argues for limits on how long a loop continues attempting a task, so a failure does not become indefinite execution.[3]