Zack Proser is an AI engineer on WorkOS’s applied AI team. He builds retrieval systems, agent harnesses and evaluations, and publishes detailed accounts of the engineering and interface decisions behind them.[1]
Start with Applied AI: Three Learnings from Shipping →
Links
Website
GitHub LinkedIn 
X · zackproser AI Engineer, Applied AI · WorkOS
Career
Proser’s public timeline places him in applied AI at WorkOS from 2025, following developer-education work there in 2024–2025. Earlier roles include staff developer advocacy at Pinecone, technical leadership at Gruntwork and senior engineering at Cloudflare. His work spans databases, APIs, frontends and the infrastructure code that deploys them.[1]
Contributions
His showcase write-up follows work from capture through planning, execution and delivery. The examples come from internal AI tooling at WorkOS, where an incomplete handoff can leave the user doing the final coordination manually. His public tutorials and long-form field notes make the implementation choices inspectable, including retrieval, evaluation and the harness surrounding an agent.[2][3]
Ideas
In his WorkOS Applied AI Showcase segment, Proser draws three lessons from shipping internal tools: the interface matters, the workflow needs to reach completion, and teams need to keep reconsidering what becomes possible. He contrasts a capable technical stack with the experience through which a person actually uses it. This makes application design part of agent engineering rather than decoration added after the model works.[2]
Tools, writing & talks
Writing, research & talks
Start here
Applied AI: Three Learnings from Shippingzackproser.comA presentation about interfaces, complete workflows and lessons from WorkOS’s internal AI tools.
Start with the showcase segment for practical product lessons before choosing a technical tutorial. It is useful if you have built a prototype and need to make the resulting workflow useful to other people.
Presenter[2]WorkOS engineers walk through running coding agents day to day: voice input, goals with measurable stop conditions, worktree isolation, hooks for lint and tests, second-model diff review, and scheduled automation, built around a faked-test failure.
Catalog workOriginal source- Zack Proser videoszackproser.com
Tutorials, talks and live coding on retrieval systems, agent workflows and applied AI.
Creator[3]