Bolder Apps je agencija za razvoj mobilnih in spletnih aplikacij, osredotočena na umetno inteligenco, s sedežem v Miamiju, ki sodeluje z zagonskimi podjetji in podjetji po vsem svetu. Njena ekipa podpira celoten življenjski cikel izdelka – od strategije in oblikovanja UI/UX do izvornega in večplatformskega razvoja mobilnih aplikacij z uporabo tehnologij Flutter, Swift in Kotlin ter spletnega razvoja z Reactom, Node.js in Laravelom. Agencija zagotavlja tudi zasnovo zaledne arhitekture, optimizacijo rasti, začasno okrepitev ekip, stalno podporo in integracije umetne inteligence. Bolder Apps deluje na področjih finančne tehnologije, zdravstva, e-trgovine, družbenih medijev in življenjskega sloga, zato je zanimiva za kandidate, ki jih zanimajo produktno inženirstvo, delo s strankami in praktični razvoj rešitev z umetno inteligenco.
Founding AI Agent Harness Engineer
Build the runtime, evaluations, prompts, and safeguards for an AI coworker startup. Ship dependable tool-using agents for real users across the United States.
Opis
- Develop the agent runtime, managing context, tool selection, approval pauses, and decisions about when the agent should act, ask, or remain quiet.
- Create and continually refine system prompts.
- Lead the evaluation and regression harness, testing prompt changes with the rigor used for code.
- Ensure long-running tasks recover from interruptions, retries, and unreliable APIs.
- Build safeguards, approval steps for high-risk actions, and strict separation of customer data.
- Work directly with the founder and early users, turning their feedback into shipped product behavior.
Zahteve
- Have launched production LLM agents that use tools with real users; demos and prompt wrappers alone do not qualify.
- Show sound judgment about agent initiative, backed by test results.
- Have practical experience building evaluations and regression suites for LLM behavior.
- Understand safe tool execution, prompt injection, and risks to data in multi-tenant systems.
- Bring solid Python or TypeScript backend skills, including asynchronous jobs, retries, and state management.
- Be able to take an ambiguous problem from prototype to production and explain the tradeoffs.
- Communicate clearly in writing and conversation.
- Bonus: experience with MCP or integrations for Slack, Gmail, Notion, or Linear.
- Bonus: familiarity with evaluation and tracing tools such as Braintrust, Langfuse, or LangSmith.
- Bonus: experience as an early engineer at a startup.
Ugodnosti
- Competitive compensation and equity.
- A founding role with ownership of the agent harness and influence over product behavior.
- Direct access to the founder and early users, without intermediary layers.
- Remote work from anywhere in the United States, ideally with working hours that overlap Pacific time.
- Access to the tools and models needed to do the work well.
- A small, experienced team that trusts you to manage your own work.
- A path to full-time employment if the arrangement proves successful for both sides.