I joined Tobias Macey on the Data Engineering Podcast this week to talk about how agents are (and aren't) changing the world of databases and data management. There's a whole lot of new and exciting things going on in this space, many driven or accelerated by AI, but also a lot of things that haven't changed and likely won't change in the near future. I spend a lot of my time thinking about this topic of what's changing and what's not, and it was fun to explore it over an hour on Tobias's podcast. https://lnkd.in/gJBsQR5w
Ravi Thakur
2w
Marc, Great discussion with Tobias! You nicely articulated how agents are changing the infrastructure game while keeping core patterns intact. The key takeaways for me: • 𝐒𝐞𝐫𝐯𝐞𝐫𝐥𝐞𝐬𝐬 𝐍𝐞𝐜𝐞𝐬𝐬𝐢𝐭𝐲: Agentic workloads accelerate the demand for serverless, elastic databases (like Aurora and DSQL) to handle spiky usage and rapid provisioning. • 𝐒𝐭𝐚𝐭𝐞 𝐌𝐚𝐧𝐚𝐠𝐞𝐦𝐞𝐧𝐭: The runtime challenge of balancing stateless LLM calls with durable, long-running agent sessions is addressed by primitives like Bedrock AgentCore runtime and its VM isolation model. • 𝐃𝐮𝐫𝐚𝐛𝐥𝐞 𝐏𝐚𝐭𝐭𝐞𝐫𝐧𝐬: Despite the hype, you correctly noted that relational databases and SQL remain foundational and critical for data with specific, durable shapes
Cloud Solutions Architect @ Ciena | AWS Community Builder
2wThere is certainly a lot of promise in this space for Agentic workflows.