Harness Engineer

AI Fabrik builds an edge inference delivery network for high-performance tokens, with faster time-to-market from grid to tokens. Our mission is to build the inference infrastructure we wished every enterprise already had — close to users, close to the cloud, and extremely resilient for real‑time workloads. We are builders, architects, engineers, and researchers with hands‑on experience in real‑world AI deployment in production, and decades of data center experience that taught us exactly what needs to change. AI Fabrik was incubated inside Gruve and backed by Mayfield, Xora (Temasek), Acclimate Ventures, Cisco Investments — existing investors from Gruve who followed us into this new chapter. We are deploying five initial production sites, with the first one coming online in July 2026. About the Role We’re hiring someone to own our engineering harness — the tooling and AI workflows that run across the full delivery pipeline, from how requirements get written to how releases go out. The job is practical: work out where teams are losing time between ideas and shipped software, and build the infrastructure that fixes it. AI agents are the main lever. The measure of success is whether teams are actually shipping faster and with fewer manual handoffs, not whether the agents themselves are impressive. Key Responsibilities Identify bottlenecks across the full software delivery lifecycle — requirements refinement, design review, implementation, testing, and production rollout — and build AI‑assisted harnesses that remove them; measure impact in concrete terms: cycle time, lead time, defect rate, and deployment frequency Build harnesses that help teams move faster from idea to implementation, including AI‑assisted requirement elaboration, automated design review, specification validation, and artefact generation Design and maintain the agent environments that support day‑to‑day development work: code generation, automated refactoring, test authoring, and documentation; define the constraints, context files, and tool permissions that keep agents accurate and scoped Build agent‑assisted release workflows, automated pre‑deployment checklists, rollback triggers, and progressive rollout controls that reduce manual effort and release risk Define what good output looks like at each stage of the pipeline and enforce it automatically through evaluation datasets, grading rubrics, and CI/CD‑integrated quality gates Instrument every harness component with structured logging, tracing, and quality metrics; use that data to identify where agents underperform, where humans are compensating, and where the next productivity gain is Implement the permission boundaries, guardrails, and human‑in‑the‑loop checkpoints that keep agents operating within safe bounds without sacrificing delivery speed Apply token optimisation, caching strategies, model tiering, and budget controls to ensure productivity gains are not eroded by runaway LLM costs Basic Qualifications 3+ years of professional software engineering experience, with a strong background in developer tooling, platform engineering, or DevOps Demonstrated focus on software delivery performance — experience measuring and improving cycle time, deployment frequency, or release quality Proficiency in at least one mainstream programming language such as Python or Go Hands‑on experience integrating with at least one major LLM API (OpenAI, Anthropic, or Google) Solid understanding of production reliability patterns: retry logic, circuit breakers, rate limiting, timeout handling, and graceful degradation Experience with CI/CD pipeline design and the full software delivery lifecycle Experience with monitoring, logging, and observability of production systems Strong written communication skills — able to document architectural decisions, constraints, and harness behaviour clearly for engineering teams Preferred Qualifications Experience with context engineering: token budgeting, retrieval‑augmented generation (RAG), and context assembly across multi‑step workflows Familiarity with agent frameworks (LangGraph, CrewAI, Claude Agent SDK, or equivalent) and agent configuration patterns Knowledge of LLM security practices, including prompt injection defence, PII handling, and output filtering Familiarity with the Model Context Protocol (MCP) and experience integrating MCP servers into development toolchains Background in multi‑agent orchestration: coordination patterns, shared state management, and parallel execution Familiarity with responsible AI practices and awareness of safety, compliance, or governance considerations in production AI systems Contributions to open‑source developer tooling, harness infrastructure, or AI engineering projects Why AI Fabrik At AI Fabrik, we hire for impact. We want those who challenge how inference infrastructure is built and who excel at delivering it in production. We are builders, architects, engineers, and researchers. We move fast, work with rigor, and care deeply about what runs in the real world. We are committed to building a diverse and inclusive team. AI Fabrik is an equal opportunity employer. We welcome applicants from all backgrounds and thank all who apply; however, only those selected for an interview will be contacted. Please note that this is an onsite position based out of AI Fabrik’s Redwood City, California office. #J-18808-Ljbffr AI Fabrik