Session
Code That Survives the Crash: Durable Task Scheduler for Workflows and AI Agents
Every system that runs long enough eventually crashes in the middle of something important. Maybe it was charging a customer, provisioning a machine, or waiting for someone to approve a change. When it comes back up, you're left asking what actually happened and where to continue from.
Durable execution is built for exactly this. You write your logic as normal code, and if the process dies halfway through, it continues from the last step that finished instead of starting over or leaving things half done. We'll see how to build that kind of workflow with Durable Task, including timers and waiting for a human to respond, and run it on Azure Durable Task Scheduler, a managed backend that keeps the state while your workers run wherever you like.
And this isn't only a problem for classic code. AI agents are long-running work too, and they fail in the same places. A lot of agent frameworks save the chat history and call that "durable". It isn't. Your agent can remember the whole conversation and still have no idea whether the payment it was making went through. We'll see what it takes to make agents and multi-agent workflows pick up after a crash and finish the job exactly once, including waiting for a human without keeping anything running.
None of this is tied to one language. The problem is the same whatever you write your code in, and Durable Task Scheduler has SDKs with the same idioms across many languages and AI frameworks, so what you learn here carries over to your own stack. We'll finish with the things that bite people in production: keeping workflows deterministic, making side effects safe to retry, and deploying new code while old work is still running.
Level: Intermediate (300). Preferred length: 50 min + 10 min Q&A. Demo-driven; demos in C# and Python, runnable on the local Durable Task Scheduler emulator (no Wi-Fi dependency). Audience: developers and architects building long-running workflows or AI agents, in any language.
Tamir Dresher
Principal Software Engineer, Microsoft Azure Serverless (Durable Functions)
Tel Aviv, Israel
Links
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top