The Course on Designing Agent-Based AI Systems offers practical understanding on how to create reliable, scalable, and production-ready AI applications based on large language models. The program covers key architectural approaches, engineering practices, and modern tools necessary for building enterprise intelligent systems.
Fundamentals of Building RAG Systems
The first module is devoted to the fundamentals of Retrieval-Augmented Generation and the structure of knowledge retrieval systems.
Basic Elements of RAG
- creation and use of embeddings;
- dividing documents into fragments;
- vector search: lexical, semantic, and hybrid methods;
- working with vector databases and FAISS and HNSW algorithms.
Industrial Architecture of RAG
- multi-tenant search and metadata filtering;
- access control;
- incremental indexing and keeping data up-to-date;
- performance analysis and bottleneck elimination.
Context Engineering in LLM Applications
In this section, you will learn how to correctly form context for large models and manage AI system memory.
Memory and Context Management
- architecture of short-term and long-term memory;
- storing user preferences;
- token optimization, compression, and summarization;
- selective data loading and preventing the use of obsolete information.
Designing Agent-Based AI Systems
This major module is dedicated to creating the architecture and logic of intelligent agents.
Architecture and Interaction of Agents
- main design patterns of AI agents;
- execution environment organization (Harness Engineering);
- integration with external services, tools, and use of function calls;
- working with MCP protocol and ensuring tool compatibility.
Reliability and Autonomy
- error handling and failover mechanisms;
- interaction between multiple agents;
- building long-term autonomous processes;
- state preservation and human involvement in the decision-making loop.
LLMOps and Industrial Operation
This module covers engineering practices necessary for reliable operation of AI applications in production.
Monitoring, Quality Control, and Model Versioning
- model monitoring and tracing;
- error diagnosis and quality testing of responses;
- automated verification pipelines;
- model lifecycle and version management.
Cost and Performance Optimization
- token consumption management;
- request and response caching;
- model routing and model ensembles;
- ensuring reliability, security, and guardrails.
Practical Design of Complex AI Solutions
The final block is entirely dedicated to practice. You will develop an industrial AI agent for programming, create key components of the system, and explore real-world application scenarios of technologies.
The outcome of the training will be the skill to build scalable agent-based AI systems capable of operating stably in corporate environments and solving complex applied tasks using modern artificial intelligence technologies.
