Melanie Warrick is senior staff developer advocate for AI at Temporal and a cofounder of Fight Health Insurance. Her engineering and teaching work focuses on AI applications that remain useful through failures and real-world constraints.[1][2]
Start with Temporal & AI and durable-agent demos →
Education
Links
Website
GitHub 
X · nyghtowl Senior Staff Developer Advocate, AI · Temporal
Cofounder · Fight Health Insurance
Career
Warrick began in film at USC, then pursued an MBA and consulting before learning Python at Hackbright Academy. She worked in data science and machine learning at Change.org and was a founding deep-learning engineer at Skymind. Later roles included data engineering and developer relations at Google and engineering leadership at Tailscale and Virta Health. That experience informs her current developer advocacy at Temporal.[1]
Contributions
Show what survives a failure
Warrick’s Temporal demonstrations examine durable execution: what state survives a crash, where a workflow stopped and how it continues. Her Temporal & AI video series separates an agent’s remembered information from its execution state. That lets long-running applications resume accurately after process or network failures.[2]
Evaluate the application’s actual task
With Holden Karau, Warrick cofounded Fight Health Insurance to help people draft appeals of denied claims. She publishes the engineering behind it, including model evaluation for appeal letters and its MCP server. Her Vibe Check series builds AI applications live with guests, exposing debugging that a polished demonstration can hide.[2][1]
Tools, writing & talks
Writing, research & talks
Start here
Temporal & AI and durable-agent demosnyghtowl.comWarrick’s demonstrations distinguish stored agent knowledge from the state needed to resume execution.
Start with her durable-agent demonstrations to see the difference between remembering facts and resuming unfinished work. Some experience with a tool-calling agent makes the failure-and-recovery examples easier to follow.
Creator and presenter[2]Temporal's Melanie Warrick shows how to build human-in-the-loop agents with durable execution, using wait conditions and signals with Google ADK and LangGraph, and demonstrates recovery after killing the worker mid-approval.
Catalog workOriginal source- Evaluating models on appeal lettersnyghtowl.com
An account of evaluating model output for the concrete task of drafting health-insurance appeals.
Author[3]