Course Objective & Scope
In multi-step tasks the context window fills up quickly, tool outputs create noise and the agent forgets critical decisions from earlier sessions. This programme covers working memory, episodic, semantic and procedural memory layers alongside context engineering techniques at an engineering depth.
Participants implement summarisation, selective pruning, tool-output filtering and vector/key-value memory stores using LangGraph Store and Mem0, producing a KVKK- and privacy-compliant agent memory architecture and checklist for a case adapted to their organisation.
Who Should Attend?
- AI engineers building long-horizon or personalised agents
- Backend developers designing LLM assistant and automation architectures
- Technical leads looking to reduce context cost and latency
- Prerequisite: Python, LLM APIs and basic agent development experience
Key Takeaways
- Architect working memory alongside episodic, semantic and procedural memory layers
- Apply summarisation, pruning and compaction strategies that prevent context overflow
- Manage large tool outputs and document chunks without polluting the context
- Persist cross-session user and task memory safely
- Audit memory retrieval quality, token cost and privacy risks
Curriculum
Agent Memory Types and Context Limits
- Working, episodic, semantic and procedural memory distinctions
- 'Lost in the middle' and cost impact in long context windows
Context Engineering and Compaction
- Sliding windows, hierarchical summarisation and selective message pruning
- Managing large tool calling outputs via references and summaries
Persistent Memory Architecture
- LangGraph Store, Mem0 and hybrid vector/relational stores
- Memory extraction, updates and conflict resolution
Prompt Caching and Cost Optimisation
- Using prompt caching in OpenAI, Claude and Gemini
- Stable prefix design to cut latency and token spend
Privacy, Right to Erasure and Security
- Tenant/user isolation, TTL and data deletion under KVKK
- Defending against memory poisoning attacks
Other Courses in This Category
-
Multi-agent architectures that split complex work across specialised agents: supervisor, hierarchical and swarm patterns, comparing LangGraph, CrewAI and AutoGen, the A2A protocol and shared state management.
- Live Course
- 5 modules
-
Human-in-the-loop agent architecture that preserves human oversight on critical actions: interrupt/resume flows, risk-based approval thresholds, escalation queues and agents that improve from human feedback.
- Live Course
- 5 modules
-
Connecting enterprise databases, APIs and internal systems to LLMs and agents cleanly and securely with Model Context Protocol (MCP): MCP server/client development, authorisation and enterprise integration architecture.
- Live Course
- 5 modules
-
Turning institutional knowledge, procedures and tools into modular, reusable agent skills: SKILL.md structure, progressive disclosure, templates, scripts and enterprise skill library governance.
- Live Course
- 5 modules
-
End-to-end and step-level evaluation of multi-step agent systems: trajectory evals, tool calling accuracy, task completion metrics, simulated user/environment testing and CI/CD eval pipelines.
- Live Course
- 5 modules
-
Securing tool-using agents that act on external systems: indirect prompt injection defence, delegated user authorisation with OAuth/OBO, sandbox isolation, guardrails and OWASP Agentic security controls.
- Live Course
- 5 modules
Get in touch
