Agentic AI: A Hands-On Workshop From Google

Join MIT Machine Intelligence and Machine Operations (MIMO) and Google for an immersive deep dive into Agentic AI - From Theory to Practice on February 28th in Cambridge, MA. This comprehensive workshop combines insights with hands-on experience, featuring in-depth exploration of reasoning models, Agentic AI, and the frontiers of AI research. Participants will build their own AI agents and multi-agent systems and explore leading tools like Veo, Flow and Nano Banana.

Workshop Outcomes:

• Advanced AI Understanding: Master the latest developments in agentic and generative AI, including reasoning models, agentic models, and multi-agent orchestration through live demonstrations and hands-on experience.
• Practical Agent Development: Gain direct experience building and implementing agentic systems through an extensive hands-on workshop, learning essential patterns and real-world applications.
• Direct Expert Access: Engage with Google professionals throughout the workshop, providing unique opportunities for specific guidance and insights.
• Comprehensive Learning Experience: Participate in a well-structured program that progresses from foundational GenAI concepts through advanced topics like embodied intelligence, culminating in open discussions that address your specific implementation questions.

Agenda

• 10 - 10:45 am: Generative AI Overview
  • Evolution of Transformers over time (Encoder & Decoder -> Decoder only)  
  • Expansion of Generative AI to new modalities: video (Veo), Music (Lyria), 
  • High-level overview of emerging research areas: test-time learning, RL post training, linear attenion layers, and more
• 10:45 - 11:30 am Hands on Session: Making a short film! 
  • Using Flow, Veo, Nano Banana, and other Generative AI tools to make an Ad (we'll look at free credit tier)
• 11:30 am - 12:15 pm Theory Session: Agents
  • So what? Aren't agents just a prompt + tools + LLM?
  • The role of MCP / A2A and other protocols
  • What does an "agent stack" look like? 
  • GCP specific agent services 
• 12:15 pm - 1 pm Lunch
• 1:00 PM - 1:45 PM: Hands-on Session: Intro to Agents! 
  • Build your first agents in cloud environment
  • Agent orchestrator (ADK), tools, deployment
• 1:45 pm - 2:15 pm Theory Session: Agents 102 
  • Dealing with LLM context limitations with memory 
  • Testing agentic systems
  • Evaluation patterns (offline & online)
• 2:15 pm - 3:00 pm Hands on Session: Build more complex agents
  • Multi-agents (e.g. a portfolio analyzer)
  • Memory
  • Evals 
  • Data with MCP
• 3 - 3:30 pm: The Frontier!
  • World models! Genie 3, Waymo Simulator, etc. A multi-modal future 
  • Unsolved problems:
    • giving LLMs "memory" beyond weight matrices
    • Hallucinations & LLM confidence 

Contact mimo-workshop@mit.edu with any questions.

Machine Learning for Manufacturing and Operations (MIMO) | Website | View More Events

Contact the organizers