Principal Engineer, AI Systems Our Client About the role Our client is a dual-use, AI-native production technology company focused on complex industrial projects. Its platform transforms fragmented engineering, planning, material, and execution data into a living production model that connects what needs to be built, how it should be built, what it depends on, and what is ready to execute now. Starting in U.S. shipbuilding, the platform is designed to scale across complex manufacturing, energy, and other large physical projects. The company currently works with a major U.S. shipbuilder and is scaling to work directly with the U.S. Navy and energy sector. What you'll do • Take ideas from zero to production. Start from a customer problem, whiteboard it, choose the architecture, build, deploy, observe, and iterate. • Build the whole application when needed. Frontend, backend, APIs, databases, auth, infrastructure, deployment, and monitoring. • Build excellent product interfaces. Turn complex production data into intuitive workflows, tables, graphs, timelines, and operational views. • Build AI-native applications. Agents, tools, prompts, structured outputs, RAG/context systems, model routing, memory, evals, and human-in-the-loop workflows. • Design agent architectures with judgment. Decide when something should be an agent, a tool, a workflow, a deterministic service, or a traditional algorithm, and avoid making everything an agent. • Connect AI to real software. Agents that query the production graph, inspect work packages, evaluate readiness, call deterministic algorithms, and return auditable results. • Own application architecture. Service boundaries, data ownership, sync vs. async execution, graph vs. relational storage, event architectures, and system interfaces. • Tame messy industrial data. Turn Excel files, ERP exports, schedules, BOMs, drawings, and incomplete customer data into working workflows. • Build around our client's production model. Parts, assemblies, systems, PWBS, MBOM lines, work packages, tasks, constraints, dependencies, resources, and execution state. • Build graph-heavy applications. Dependency networks, critical paths, readiness states, lineage, provenance, and change propagation across very large production graphs. • Productize algorithms and research. Expose scheduling logic, readiness rules, graph algorithms, optimization, and simulation as reliable, observable, performant product capabilities, working alongside our research engineer. • Prototype fast, then harden what matters. Ship a credible new application or major workflow in days or weeks, then refactor the parts that matter into durable infrastructure with tests and clean interfaces. • Deploy and operate what you build. AWS, containers, CI/CD, logging, monitoring, security, secrets, and performance. • Debug across the full stack. Trace a wrong answer from UI to API to agent to model to database to the underlying production object. • Decide without paralysis. Make a sound 80–90% call quickly, understand the tradeoffs, and preserve optionality where uncertainty remains. • Work directly with operators. Watch how a planner or production manager works and translate it into software without a PM mediating every conversation. Who you are • Builder first. You've personally taken complex products from a blank repo or early prototype to real production use. • Truly full-stack. Comfortable across React/TypeScript, backend services and APIs (Python or similar), data models, cloud infrastructure, and deployment. • AI-native engineer. Hands-on with LLMs, agents, tool use, structured outputs, RAG, evaluation, and agent orchestration. • Strong architecture judgment. You make consequential calls on boundaries, state, APIs, data models, async and event-driven processing, and scaling, while still writing a lot of code. • Clear on where AI belongs. You separate probabilistic model behavior from deterministic computation and combine them safely. • Production-minded prototyper. You move fast without building throwaway demos, and know which shortcuts are reversible. • Data-intensive experience. You've built on complex relational, graph, temporal, or operational data, not just CRUD apps or chat interfaces. • Product instinct and autonomy. Given a capability and an imperfectly stated problem, you figure out the architecture, build it, ship it, and own the outcome. • At home at the research boundary. You turn experimental models and algorithms into reliable product features. • Excellent communicator. You move easily between researchers, engineers, shipyard operators, customers, and leadership. • Thrives early-stage. Comfortable with incomplete information, shifting priorities, thin infrastructure, and broad ownership. Experience we'd value • 7–10+ years of exceptional software engineering; scope and depth matter more than years. • Early or founding engineer at a high-growth company, ideally taking a product from prototype to meaningful production scale. • Or substantial engineering at a technically demanding organization (e.g., Tesla, Palantir, Anduril, SpaceX, OpenAI, Anthropic, Scale AI, Databricks, Stripe). Especially compelling: both a rigorous engineering org and an early-stage startup. • Evidence of things built outweighs titles: launched products, open-source work, side projects, or systems you personally owned. • Bonus: software that touches the physical world (manufacturing, robotics, autonomy, aerospace, defense, energy, logistics, construction, industrial). • Major plus: AI deployed where accuracy, provenance, reliability, security, and human review matter. Education • B.S./M.S. in Computer Science, Computer Engineering, Electrical Engineering, Applied Mathematics, or a similarly rigorous discipline preferred, not required. • Advanced degrees welcome. Demonstrated ability to build difficult systems carries far more weight than credentials, and builders with unconventional backgrounds are equally considered.