From learning to leaderboard: How Cisco upskilled 1,200 engineers in agentic AI with AWS AI League
Cisco upskilled 1,200 engineers by running an AWS AI League tournament where participants built agentic AI solutions using Amazon Bedrock AgentCore and fine‑tuned models in Amazon SageMaker Studio, receiving real‑time leaderboard feedback. The competition evaluated agents on maze navigation, safety (Bedrock Guardrails), tool integration (AgentCore memory, Lambda Event Functions), and model accuracy versus a reference baseline. Engineers gained hands‑on experience that translated directly into production‑ready agent designs, demonstrating a scalable “learn‑build‑compete” pathway for enterprise AI skill development.
⚡ Key Takeaways
- 1,200 Cisco engineers competed in the 2026 AWS AI League tournament.
- Agents were built with Amazon Bedrock AgentCore, leveraging AgentCore memory and AWS Lambda Event Functions.
- Evaluation metrics included time efficiency, accuracy, agent planning, and token‑consumption efficiency.
- Participants customized supervisor and sub‑agent models and applied Amazon Bedrock Guardrails for safety.
- Model fine‑tuning was performed in Amazon SageMaker Studio using the latest training recipes.
Providing a gamified, hands‑on competition accelerates enterprise AI adoption by turning abstract demos into concrete, production‑grade agent implementations, reducing the learning curve for large engineering teams.
✅ Practical Steps
- Attend the 2‑hour AWS AI League “Learn” workshop to get introduced to Amazon Bedrock AgentCore and SageMaker Studio.
- Build an agent that navigates the maze grid, integrates AgentCore memory and Lambda Event Functions, and submit it to the live leaderboard for automated evaluation.
- Fine‑tune a model in SageMaker Studio using the provided training recipes, then submit the customized model for scoring against the reference baseline.
Want the full story? Read the original article.
Read on AWS ML Blog ↗