Workshop

Introduction To AI Agent Memory

Richmond Alake

Coined Memory Engineering | AI @Oracle

Your agents cost too much and deliver too little. Be the one who fixes both.

AI agents are the dominant form factor of software today and for the foreseeable future. Every company running them is asking two questions: why does this cost so much, and what has it changed?

Someone in your company will answer both. If you're an AI Architect, Forward Deployed Engineer, AI Product Manager, or Solutions Architect, it should be you.

The answer is agent memory: how agents retain, reuse, recall, and refine the information they pull from their environment and execution traces. It is what makes agents reliable, believable, and capable.

It is also your biggest tokenomics lever.

Managing information inside and outside the LLM's context window (through smarter retrieval, storage, and refinement) cuts an agent's operational cost while lifting performance.

Lower cost and better results from the same work. In one session you get:

  • The vocabulary to make the cost and impact case to leadership

  • The mental models for memory in Assistant, Workflow, and Deep Research applications

  • The architecture map showing where the savings hide in your own agents

  • The implementation steps to building an agent memory management system.

What you’ll learn

Go from treating memory as a feature to designing it as infrastructure, and become the one who cuts agent costs and owns the impact.

  • Work from a precise definition: memory as infrastructure that is external to the model, persistent, and structured

  • See where prompt engineering and context engineering stop, and why memory is what decides an agent's cost and quality

  • Use a simple test for what counts as memory versus retrieval, state, or cache

  • See how popular AI products such as ChatGPT and Claude use memory in their consumer-facing applications

  • Trace short-term memory (the context window) and long-term memory (external stores) through live product and implementation walkthroughs

  • Implement examples of short- and long-term memory in an agentic application using open-source tools and SDKs from frontier labs

  • Implement the memory types in an agentic application and leave with a reference architecture for your own systems

  • Follow the data-to-memory pipeline: extraction, consolidation, and refinement of raw interactions into durable knowledge

  • Compare storage and retrieval options (vector, graph, relational) and when each one fits

  • Map the full lifecycle: what gets written, consolidated, retrieved, isolated per user, and forgotten

  • See why retrieval (RAG) is one read operation, a single stage of the job, not the job itself

  • Understand every component of an agent memory system, concept to implementation: embedding models, retrieval mechanisms, and memory stores

  • Apply prompt caching, semantic caching, compaction, and consolidation to bring down spend per turn

  • Decide what belongs in the context window and what belongs in long-term memory, and what each choice costs

  • Know which lever to pull first on the agent you already have

Workshop agenda

  • What is agent memory?

    Agent memory from first principles: what it borrows from neuroscience, a concrete definition and why it drives performance and cost, the taxonomy across three lenses, and a reference architecture.

  • Implementing memory-aware agents

    Build the memory types into a working agentic application: a detailed implementation with walkthroughs of the key code, plus the core components of an agent memory system in practice.

  • Cutting agent costs: the tokenomics toolkit

    Implement context compaction and reduction, memory evolution and consolidation, and the retrieval strategies that bring an agent's operational costs down while performance goes up.

  • Application modes and who owns agent memory

    Assistant, Workflow, and Deep Research mapped to the memory each demands, and the ownership gap: who is accountable for what your agent writes, recalls, and forgets, and why that should be you.

  • Live Q&A and your cost and impact roadmap

    Open questions answered live, then the path onward: the reference architecture recap and where to look first for savings in your own agents.

Learn directly from Richmond

Richmond Alake

Richmond Alake

Contact

Built the agent memory course with Andrew Ng's team. Trains enterprise AI teams.

See all products from ascxend

Who this workshop is for

  • AI Architects and Solutions Architects who design agent stacks, get asked why they cost so much, and want an answer they can defend.
  • Forward Deployed Engineers and AI Engineers whose agents hit a wall in production: forgetting users, repeating work, burning budget.
  • AI Product Managers and technical leads who own the outcome and the budget, and want to be the one credited for savings and impact.

What's included

Richmond Alake

Live sessions

Learn directly from Richmond Alake in a real-time, interactive format.

Lifetime access

Go back to course content and recordings whenever you need to.

Certificate of completion

Share your new memory engineering skills with your employer on LinkedIn

Maven Guarantee

Your purchase is backed by the Maven Guarantee.

Frequently asked questions

Maven for Teams

Reimbursement

Get your company to pay

Everything L&D needs: email template, receipts, and certificate of completion.

Get reimbursed

Team discount

Learn with your teammates

Save 20%+ when 2 or more teammates enroll in the same cohort.

Save 20%+ with a team

Private cohort

Run a cohort for your org

A dedicated cohort with a custom schedule and curriculum, tailored to your team.

Book a private cohort

£300

GBP

Oct 1
·

12–3pm EDT

Enroll