-
In Building AI Products—Part I: Back-end Architecture Phil Calçado shares invaluable lessons from building and scaling the back-end of their AI PaaS product.
A surprising takeaway? Traditional microservices didn’t cut it for AI agents, which thrive on stateful operations and rich context. Instead, Calçado and his team found success with an object-oriented design, which simplified implementation and state management.
Another interesting design decision is the natural language event handling using proposition-based retrieval, transforming unstructured messages into structured, actionable data for efficient processing.
It is an exciting time as we are collectively figuring out how existing patterns can be adapted and new patterns emerge.
-
The current (r)evolution in AI, driven by Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and other advancements, has been captivating to follow. One invaluable source for staying updated has been Simon Willison’s link blog. I’ve been closely following it for a while, appreciating the updates enriched with thoughtful commentary and context. While there are numerous […] read more
