Battle-Tested Guide to Building Multi-Agent Systems in 2026

Hosted by Valentine Zavadskiy

Tue, Jan 20, 2026

5:00 PM UTC (1 hour 30 minutes)

Virtual (Zoom)

Free to join

Invite your network

Go deeper with a course

AI for Developers: From Vibe Coding to Multi-Agent Systems
Valentine Zavadskiy and Olga Lavrentieva
View syllabus

What you'll learn

Ecosystem & Language Trade-offs

We'll compare Python’s rich research ecosystem and type safety and latency advantages of TypeScript in backend systems

Architectural Trade-offs: SDKs vs. Frameworks

We examine when to use direct SDKs or LiteLLM for low overhead, and when frameworks are needed for complex orchestration

The Anatomy of a Modern Agent

A breakdown of core components of agentic systems, from structured outputs and tool discovery to memory and coordination

Deep Dive: The Big Two Competitors

We compare LangChain/LangGraph and CrewAI using real code to determine which framework best fits your use case

How Agents Connect to the Real World (Action Layer & MCP)

We explore MCP and how it standardizes interactions between agents and databases, local files, and APIs

Why this topic matters

In addition to the topics above, we cover Production, Monitoring, and Deployment, moving beyond localhost to discuss production infrastructure, performance trade-offs, observability, and cloud scaling. As a result, you’ll have a clear mental model of the AI agent ecosystem and be able to make informed architectural decisions based on complexity, type safety, performance — not tool popularity.

You'll learn from

Valentine Zavadskiy

Co-founder at evy.so, former co-founder of Narrative BI & Rubyroid Labs


Co-founder at evy.so — a single proactive voice assistant for teams. It speaks, listens, and remembers context across everything you work on, helping teams stay focused and productive hands-free — whether at their desks or on the go.


Software Engineer with 15+ years of experience. Serial entrepreneur with successful exit. Taught thousands of students how to code. Co-host of the tech podcast with 200+ episodes live. His current focus is on building AI solutions and exploring a unique, highly efficient way of developing software using voice.


For the past three years, Valentine has been creating commercial AI solutions where large language models (LLMs) are at the center. His journey began in a startup that developed a system allowing users to gain insights and analytics simply by talking to their data. Since then, he has worked with:

  • Different LLMs, from OpenAI to specialized models.
  • Fine-tuning models for specific business needs.
  • Designing and applying advanced patterns for effective AI interaction.


His extensive practical experience is combined with an ability to explain complex concepts in simple language and to create engaging, effective educational solutions.

Previously at

Narrative BI
Rubyroid Labs

Sign up to join this lesson

By continuing, you agree to Maven's Terms and Privacy Policy.