À propos
But we don't build software the traditional way anymore.
We ship production features at high velocity using AI-first workflows. Most days, we rarely touch code directly. We design systems, orchestrate agents, and iterate through structured prompting, evaluation loops, and tool-driven workflows.
We are looking for a true agentic developer.
Not someone "excited to learn AI."
Not someone who has tried ChatGPT a few times.
Not someone planning to grow into this role.
We expect you to show up fluent on day one.
WHAT YOU'LL DO
* Design, orchestrate, and manage multi-step agent workflows
* Build reliable development pipelines using Codex and/or Claude as core collaborators
* Architect structured prompt systems, tool chains, and feedback loops
* Debug agent behavior, not just code
* Create evaluation harnesses and guardrails for agent outputs
* Iterate rapidly on product features without manually writing most of the code
* Manage long-context reasoning and task decomposition across agents
* Use tools like Beads (or equivalent orchestration frameworks) effectively
This is not a "prompt engineer" role.
This is not a junior AI tinkerer role.
This is not a learn-as-you-go environment.
WHAT "STRONG EXPERIENCE" MEANS
You should already:
* Have built real systems where LLMs generate, refactor, and reason about production code
* Be comfortable orchestrating multiple agents with defined roles
* Understand tool invocation, structured outputs, and retrieval patterns
* Know how to prevent hallucinations with architecture, not hope
* Think in terms of feedback loops, evaluators, and self-correction
* Have experience pushing Codex or Claude beyond trivial completions
* Be familiar with agent frameworks like Beads (or equivalents)
* Have clear opinions about temperature, context management, evaluation strategy, and tool design
If you need to ramp up on agent workflows, this is not the role.
WHAT WE'RE LOOKING FOR
We know what strong agent developers look like because we already work this way.
You likely:
* Treat LLMs as deterministic components inside larger systems
* Think in abstractions, not single prompts
* Have built internal tooling around AI to increase leverage
* Move fast but design for reliability
* Care deeply about reproducibility and evaluation
* Prefer designing systems over writing boilerplate
We are not looking for:
* Traditional developers who are "AI-curious"
* Engineers who primarily write code manually
* People who believe prompting alone is the job
* Candidates who need 6 months of experimentation before becoming effective
---
ABOUT US
* Small team
* Deep traditional engineering experience
* Now operating in an AI-first development model
* Shipping fast
* High autonomy
* High standards
We optimize for leverage, not lines of code.
---
HOW TO APPLY
Instead of a cover letter, show us:
* A real agent workflow you built
* A system where AI meaningfully replaced manual coding
* Tooling you created around Codex or Claude
* Evidence that you understand evaluation and reliability
We can tell in one conversation whether you're fluent.
If you are more comfortable orchestrating intelligence than typing syntax, we should talk.
If you want, I can also provide:
* A shorter, punchier version for X / LinkedIn
* A more aggressive filter-heavy version
* A more visionary founder-style version
* A compensation-ready version for formal job boards
Contract duration of 1 to 3 months. with 30 hours per week.
Mandatory skills: Claude, Engineering & Architecture, OpenAI Codex
Compétences linguistiques
- English
Avis aux utilisateurs
Cette offre provient d’une plateforme partenaire de TieTalent. Cliquez sur « Postuler maintenant » pour soumettre votre candidature directement sur leur site.