Most companies already have developers. The question is whether the team knows how to work with AI without turning the repo into soup.
This role is for someone who can sit with an existing engineering team and make them better. Some developers will be skeptical. Some will be fully AI-pilled. Most will be uneven across tools, review habits, test discipline, deployment patterns, and comfort with agents touching real code. You need enough taste to meet each person where they are without lowering the bar.
We want a practitioner with opinions earned from shipping. Claude Code, Codex, Cursor, CI, evals, preview environments, deployment pipelines, secrets, observability, infrastructure, and the small rituals that keep AI-assisted development from becoming chaos. The job is to teach, configure, coach, and build the operating system around the team.
What you will do
- Assess how a client dev team works today: repos, review loops, CI/CD, deploys, environments, permissions, incident history, and AI usage.
- Design AI-assisted development workflows that fit the team's stack and risk tolerance.
- Coach engineers across the adoption curve, from skeptical seniors to people who want agents writing every line.
- Configure tools, prompts, guardrails, evals, tests, and review practices that make AI output safe enough to ship.
- Improve the surrounding delivery system: DevOps, infrastructure, deployment, observability, documentation, and handoff routines.
What we need
- You have shipped production software and helped other developers get better.
- You use AI coding tools heavily enough to know their real failure modes.
- You understand modern development infrastructure: GitHub, CI/CD, cloud services, environments, secrets, logging, and deployment tradeoffs.
- You can teach without becoming condescending or turning every conversation into a tools debate.
- You can distinguish a team problem, a process problem, and a tooling problem under real delivery pressure.
Strong signals
If any of these describe you, the conversation will move quickly.
- You have introduced AI development workflows to a team that did not all start in the same place.
- You have strong opinions about where agents belong in the SDLC and where they should still be watched closely.
- You can make a skeptical engineer faster without asking them to become a different person.
- Your own setup has custom commands, scripts, templates, checks, or harnesses because default workflows were too leaky.