← Back to blog
Elliot’s talk at the Effect conference in San Francisco

Our founding engineer Elliot Dauber shares his experience using Effect to build reliable AI systems at Markprompt in his first conference talk.

Our founding engineer Elliot Dauber did his first-ever talk, and it was awesome! He talked about Effect, the TypeScript library we are using to operate our agent infrastructure for customer support. Elliot had no Effect experience prior to joining Markprompt, but quickly ramped up.

Here are the highlights:

  1. Effect shines when building complex systems. While LLM providers like OpenAI and Anthropic are powerful, their performance characteristics can vary. He unpacked how Effect enables us to handle this inherent unpredictability with fallback mechanisms, retries, and granular error handling, forming the backbone of our highly reliable, always-available AI systems.

  2. Effect’s composable architecture enables us to build modular systems that snap together like LEGO blocks. From integrating with platforms like Salesforce and Zendesk to orchestrating multi-step agentic workflows using various LLM models, composability means we move fast without breaking things.

  3. Dependency injection allows us to model failure scenarios in ways most systems can’t. By simulating everything from small hiccups to full API outages, it is our internal playbook to anticipate problems before they happen. So, when things go wrong, we don’t scramble, we execute and tighten up.

  4. Effect is so much fun to work with, and brings out the best in us engineers. Building software systems for large enterprises is complex. But instead of focusing on piling up this complexity and debugging issues, Effect allows us to spend most of our time building features, in a way that just feels right.

Full talk: https://youtu.be/jsTjgB5WUFE?si=ZFR1lCwAXcfIcXRA .