Richmond Alake is Director of AI Developer Experience at Oracle and creator of MemoRizz, an open-source Python library that gives AI agents persistent, searchable memory. His work on memory engineering addresses a central limitation of agentic software: systems cannot learn from previous interactions, remember user preferences, or avoid repeating mistakes without carefully designed mechanisms for storing and retrieving experience.
Alake studied software engineering at Kingston University and earned a master’s degree in computer vision, machine learning, and robotics at the University of Surrey. His early work applied deep learning to pose estimation, gesture recognition, facial-feature detection, and image segmentation. He taught computer vision through O’Reilly and wrote technical articles for NVIDIA about neural networks, computer vision, research literacy, and data storytelling.
He joined MongoDB in December 2023 and became an AI/ML staff developer advocate, helping developers build applications with vector search, retrieval-augmented generation, and document databases. As agents acquired tools and multistep workflows, he concentrated on their missing infrastructure: durable memory. His writing on memory-augmented agents and AI Engineer World’s Fair 2025 presentation developed practical architectures for systems that retain context across interactions.
Alake joined Oracle in 2026. With Nacho Martínez, he created a DeepLearning.AI course on memory-aware agents, extending that work into database-backed architectures for persistent agent behavior.
- Memory as application architecture. Useful agent memory requires a complete lifecycle: extracting information, storing it, retrieving relevant details, incorporating them into context, updating records, and selectively forgetting. Larger context windows do not eliminate the need for disciplined retrieval.
- Different memories for different jobs. Alake distinguishes conversational history, user and entity facts, personas, working state, episodic experience, procedural knowledge, and workflow outcomes. Agents can retrieve tools only when necessary and preserve failed steps to improve later executions.
- MemoRizz. His library implements persistent agents, semantic retrieval, conversation storage, forgetting controls, and integrations spanning MongoDB, Oracle, and local workflows.
- Database-native agent memory. Alake coauthored research on Oracle Agent Memory describing memory scoped across users, agents, and conversation threads, with evaluation based on retrieval quality, recall, latency, token consumption, and downstream task performance.
