Course Objective & Scope
Agents that fail to remember user preferences, past decisions and learned rules across sessions start from scratch on every task, while dumping the entire history into the context window drives up cost and degrades reasoning quality. This course covers how to architect a layered memory system for production agents.
Hands-on labs cover working memory management, conversation summarisation, semantic and episodic storage using pgvector, Redis and LangGraph Store/Mem0 patterns, resolving conflicting facts and implementing GDPR/KVKK-compliant deletion. Participants leave with a cross-session memory-enabled agent tailored to an internal use case.
Who Should Attend?
- AI engineers building personalised assistants and long-lived agents
- Backend and platform developers designing data and memory layers
- Data scientists integrating RAG with agent memory structures
- Prerequisite: Python, embeddings and basic agent development experience
Key Takeaways
- Model working, episodic, semantic and procedural memory as distinct architectural layers
- Build dynamic trimming and hierarchical summarisation flows that prevent context window overflow
- Write background memory pipelines that extract and update user and task facts across sessions
- Combine relevance, recency and importance scoring when retrieving memories
- Apply TTL and forgetting policies to purge stale or incorrect memories
Curriculum
Agent Memory Taxonomy
- Short-term (working/thread) versus long-term (cross-thread) memory
- Use cases for episodic, semantic and procedural memory
- How dynamic agent memory differs from standard RAG
Short-Term Memory and Context Window Management
- Message history trimming, sliding windows and token budget control
- Compressing tool outputs and rolling conversation summarisation
- Checkpoint-based thread state persistence
Long-Term Memory Stores and Extraction
- Extracting structured profiles and facts from conversations
- Vector, key-value (LangGraph Store, Redis) and knowledge-graph memory stores
- Hot-path (synchronous) versus background (asynchronous) memory write patterns
Memory Consolidation, Conflict Resolution and Forgetting
- Updating, consolidating and invalidating existing facts
- Hybrid retrieval weighted by recency, relevance and importance
- Preventing memory poisoning, tenant isolation and right-to-erasure
Lab: Multi-Session Enterprise Agent
- Building an agent that remembers preferences and learns from past runs
- Evaluating memory recall accuracy and latency impact
Other Courses in This Category
-
Technical course on safely executing LLM- and agent-generated code using gVisor, Firecracker microVMs, container isolation, network/filesystem restrictions and resource quotas.
- Live Course
- 5 modules
-
Course on evaluating end-to-end task completion, trajectory quality and tool selection in AI agents, and setting up distributed tracing and live monitoring with OpenTelemetry, LangSmith, Arize Phoenix and Langfuse.
- Live Course
- 5 modules
-
Training on protecting multi-step agents against API outages, rate limits, partial failures and crashes using retries, model fallbacks, idempotency, compensating actions and durable execution with Temporal and LangGraph.
- Live Course
- 5 modules
-
Architectural design course examining prompt chaining, routing, parallelisation, orchestrator-workers, evaluator-optimizer and autonomous agent patterns, and how to pick the simplest effective pattern for each problem.
- Live Course
- 5 modules
-
Hands-on developer training on combining ReAct reasoning loops, Plan-and-Execute task planning and explicit state/checkpoint management into a single unified hybrid agent architecture.
- Live Course
- 5 modules
-
Technical course on delegating work across specialist agents using control handoffs versus the agent-as-tool pattern, covering triage routing, context filtering and conversation ownership.
- Live Course
- 5 modules
Get in touch
