Module 5 · Project 4: Production Architecture for GenAI and Agent Systems (capstone)

OBJECTIVE: design and deploy a production-ready GenAI/agent system — e.g. the multi-agent customer support or the RAG assistant — that is modular, secure, scalable and affordable, serving several teams or clients, and present it with a live demo and an operations runbook. WHAT WE COVER (in project order): • Reviewing a GenAI system architecture: what goes where • Modular design: API gateway, retrieval service, agent service, workers • Background jobs and queues for long tasks • Serving agents and LLM chains as APIs; self-hosted models (glimpse) • Authentication, authorisation and multi-tenancy • Load testing and scaling • Reliability: timeouts, retries, health checks, graceful failure • Governance: audit logs, data retention, budgets per client • Runbook, live deployment and project showcase AFTER THIS MODULE YOU CAN: • Design a GenAI system that can grow from a demo to many users • Make it secure, reliable and affordable, and prove it with load tests • Operate it with a runbook and present it professionally TOOLS: FastAPI, Docker Compose / Cloud Run, Redis + Celery (or similar), PostgreSQL, JWT auth, Locust, Langfuse, Prometheus/Grafana, vLLM (glimpse).