Senior AI Agent Engineer — Autonomous Orchestration, Self-Healing Systems & Production Reliability

We are looking for a highly experienced Senior AI Agent Infrastructure Engineer to finalize, stabilize, and strengthen an existing autonomous software-development environment: Paperclip. The system supports development of Lettuce+, our Bitcoin-native creator platform. This is not a greenfield project. Considerable engineering time, funding, and effort have already gone into the platform. Core infrastructure, orchestration, agent coordination, automated testing, and workflow automation are largely in place. We need an expert who can investigate the remaining weaknesses, repair the current design where necessary, and turn the platform into a truly autonomous, reliable, self-healing, self-validating, and continuously improving system. Existing Technology and Infrastructure Our current environment includes: - TypeScript, Node.js, Next.js, and GitHub - Codex autonomous coding agents - Claude Code review agents - Paperclip orchestration and task management - Multi-agent workflows and automated dispatch - Hetzner servers and self-hosted infrastructure - CI/CD pipelines and automated testing - Durable queues, worker leases, and recovery mechanisms - Dottie and Jarvis orchestration agents Current Challenges Progress has been blocked for several weeks on an important milestone: Wave 10, Item 15. While a significant amount of implementation is complete, the platform still has issues including: - Agents sitting idle instead of being assigned further work - Inconsistent autonomous task scheduling and dispatch - Failed jobs and expired worker leases that require manual recovery - Code-review and validation transitions that stall - Infrastructure or connectivity problems - Queue progression that does not complete autonomously - Lack of convincing evidence that the system can run unattended with reliability Our five-minute dispatcher has been extensively implemented and tested, but it has not yet proven that it can operate continuously in production without supervision. We need someone who can find the true outstanding failure points and bring the full execution cycle to completion. Primary Goal: Close the Autonomous Execution Loop 1. Autonomous Execution Enable the system to select development work automatically, dispatch coding agents, manage reviews, refill queues, and progress the roadmap without regular human involvement. 2. Self-Healing Implement reliable detection and recovery for failures, stalled agents, and expired leases; avoid duplicate execution; and restore service after interruptions or crashes. 3. Self-Verification Automate test execution, completion validation, independent review coordination, and safeguards that stop unverified work from advancing. 4. Self-Improvement Identify repeated operational problems, record root causes, make authorized corrective changes, and introduce regression coverage to prevent the same issues from returning. 5. Durable Infrastructure Ensure operational state persists, survives restarts, respects authorization boundaries, and recovers safely from infrastructure-level failures. 6. Demonstrated Unattended Operation Provide evidence of at least 24 hours of dependable unsupervised operation, including exercised recovery scenarios and documented results. Required Experience We are especially interested in engineers who have personally built and operated production autonomous coding-agent platforms. Experience with Claude Code, Codex, agent orchestration, TypeScript, Node.js, distributed systems, durable queues, CI/CD, and infrastructure reliability is strongly preferred. Experience limited to building AI chatbots is not enough for this engagement. We need someone who has engineered autonomous systems that can coordinate software work, recover from operational failures, and function reliably without constant human oversight. Budget and Engagement - Project type: Fixed price - Budget: Up to $3,500 - Initial engagement: One well-defined autonomy milestone - Payment: Milestone-based and dependent on demonstrated outcomes Our preference is to repair, extend, and make use of the existing infrastructure—not rebuild the system from the ground up. The final milestone must include documented proof of successful unattended operation. Before committing, candidates should assess whether this scope is realistically achievable within the budget and clearly communicate any constraints or limitations. Potential Future Work A successful engagement may lead to additional fixed-price work, including: - Completing development Waves 10, 11, and 12 - Redesigning Paperclip - Continuing development of Lettuce+ - Migrating applicable components to Bend 2 - Integrating Spark payment rails - Implementing planned on-chain functionality The longer-term aim is to use the autonomous system itself to accelerate these future stages, while retaining appropriate human oversight and independent security review. There may also be opportunities for occasional emergency engineering support and future consulting work.