Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Duration 35 hours
Course Outline
Advanced LangGraph Architecture
- Graph topology patterns: encompassing nodes, edges, routers, and subgraphs.
- State modelling: covering channels, message passing, and persistence.
- Comparing DAG versus cyclic flows and exploring hierarchical composition.
Performance and Optimisation
- Parallelism and concurrency patterns within Python.
- Strategies for caching, batching, tool calling, and streaming.
- Cost control mechanisms and token budgeting strategies.
Reliability Engineering
- Implementation of retries, timeouts, backoff, and circuit breaking.
- Ensuring idempotency and deduplication of steps.
- Checkpointing and recovery techniques using local or cloud-based stores.
Debugging Complex Graphs
- Step-through execution and dry runs for verification.
- State inspection and detailed event tracing.
- Replicating production issues effectively using seeds and fixtures.
Observability and Monitoring
- Structured logging and distributed tracing techniques.
- Tracking operational metrics: latency, reliability, and token usage.
- Utilising dashboards, alerts, and SLO tracking.
Deployment and Operations
- Packaging graphs as standalone services and containers.
- Managing configuration and handling secrets securely.
- CI/CD pipelines, rollout strategies, and canary deployments.
Quality, Testing, and Safety
- Unit testing, scenario testing, and automated evaluation harnesses.
- Implementing guardrails, content filtering, and PII handling.
- Conducting red teaming and chaos experiments to ensure robustness.
Summary and Next Steps
Requirements
- A solid grasp of Python and asynchronous programming principles.
- Practical experience in developing LLM applications.
- Working familiarity with fundamental LangGraph or LangChain concepts.
Target Audience
- AI platform engineers.
- DevOps specialists for AI.
- ML architects responsible for managing production LangGraph systems.