Senior Software Engineer, AI Agent Systems

Company DescriptionProDex Labs is building the AI-native operating system for American manufacturing. Our platform pairs a live discrete-event simulation engine with Dexter, our operations agent — so factory planners can model their operation, generate schedules, and test what-if scenarios in seconds instead of weeks. We're a small, well-funded seed-stage team headquartered in NYC, deployed at defense primes (L3Harris, HighCom Armor) and commercial leaders (Alimentos Prosalud, Time Manufacturing), with multiple government and commercial engagements in flight. SOC 2 Type 1, deployed in ITAR-controlled facilities, Tradewinds Awardable.Backed by Palantir Startup Fellowship, Valor Equity Partners (Antonio Gracias), The General Partnership, Sunflower Capital, and Parallel VC.Role DescriptionDexter is ProDex’s operations agent. He’s deployed inside L3Harris, HighCom Armor, and other defense and industrial primes, where he builds simulation models, configures planning runs, and helps factory operators make decisions that move millions of dollars of throughput. He is not a chatbot.We’re hiring a Senior Software Engineer to own the agentic LLM framework that powers Dexter — the orchestration layer where multiple LLMs coordinate to parse messy factory data, configure live simulations, and guide users through high-stakes planning decisions. Sub-second latency. Defense-grade reliability. No room for “looks plausible.”What you’ll own:Architecture and refinement of our multi-agent orchestration framework (LangGraph-based)Eval, regression, and reliability tooling for non-deterministic agents that take real actions on customer systemsRAG, context management, and memory systems across long-running planning workflowsThe seam between LLM reasoning and the deterministic simulation engine — where the math has to back up what the agent saysLLM cost, latency, and reliability budgets across cloud and on-prem deploymentsSetting the technical direction for how Dexter evolves over the next 18 monthsYou probably look like:Production experience building multi-LLM systems where correctness matters — not chat UIsStrong opinions on agent orchestration patterns (LangGraph or your own framework) and the failure modes of eachFluency in eval design — you know the difference between “vibes-good” and a regression-tested releaseComfortable owning a system from architecture through on-callBackground at a serious applied-AI company (UiPath, Figma AI, Adobe AI, Anthropic, an agent startup, a research lab)Bonus: experience with industrial / planning / scheduling domainsTech Lead PathCandidates with prior tech lead or staff-level experience will be considered for that scope at the offer stage. We’re a team of under 15; whoever owns this surface area will be making roadmap-level calls from day one.LogisticsNYC, in-office (5 days)$140K – $200K + meaningful equity, commensurate with experienceU.S. citizenship required (ITAR-controlled deployments)Our ProcessWe build fast, we ship fast, and we hire fast — if we identify a strong fit early you could potentially have an offer in-hand the same week.You'll meet with our CEO or COO for an initial callFrom there we'll schedule a call with one of our engineersDepending on how that goes we may ask you to perform a take-home exercise and schedule another round to go over that, or move you directly to a call with our CTOOffer-stageThis is the most technically demanding role on our engineering team. If you’ve shipped real multi-agent systems into production — not demoware, not a single-agent chat wrapper — we’d like to talk.