Teaching AI Agents Without Fine-Tuning: Context Learning for DevOps

In this 21-minute engineering session from Konvupero, Ghazi Felhi, AI Engineer at Anyshift, presents a breakthrough approach to teaching AI agents new skills without the heavy costs of fine-tuning. He walks through how Annie, Anyshift's AI SRE, monitors infrastructure, reads logs, handles alert pipelines, and runs RCAs in fast-changing DevOps environments.

⚡ Context Learning Approach: Agents execute tasks, reflect on outcomes, and build evolving cheat sheets that help Annie understand naming patterns, log behavior, and incident signals.

📊 30% Faster RCA: Benchmark results show context learning cuts root cause analysis time by approximately thirty percent compared to traditional approaches.

🧠 No Fine-Tuning Required: Learn how Annie adapts to new environments and codebases without expensive model retraining, making it practical for teams with constantly evolving infrastructure.

🙌 Huge thanks to Konvu.com for the invitation to speak at Konvupero!