What You'll BuildAs the founding AI engineer, you'll tackle problems that don't have Stack Overflow answers:Build RAG systems that understand construction terminology—teach AI the difference between 'pour concrete' and 'poor concrete'Deploy computer vision that detects safety violations from grainy phone photos taken at 6 AMCreate AI assistants that answer 'What's the status of the Stanford dorm project?' by reasoning across blueprints, contracts, RFIs, and daily photo logsDesign real-time progress tracking that works even when construction sites have terrible WiFiBuild domain-aware AI that makes construction sites safer and more efficientBuild verification systems that track equipment from PO to energization across complex supply chainsRequirementsYou are:Recent graduate from top AI program (Stanford AI Lab, MIT CSAIL, or equivalent) OR 2–3+ years building production ML systemsFocused on practical AI applications, not just research demosComfortable with the full ML stack: data processing model selection deployment monitoringAble to move quickly—you prototype in hours, not weeksMust HaveShipped at least one LLM-based application used by real usersExperience with RAG, embeddings, and vector databasesStrong Python skills plus PyTorch, TensorFlow, or JAXAbility to explain complex ML concepts to non-technical stakeholdersNice to HaveComputer vision experience (YOLO, Segment Anything, etc.)Published ML research or Kaggle competition medalsExperience with construction, manufacturing, or industrial datasetsTrack record of optimizing inference costsWhat You'll Work WithWe're flexible on the stack, but likely:LLM APIs: OpenAI, Anthropic, Gemini—multi-model approachOrchestration: LangChain, LlamaIndex, or custom frameworksVector stores: Pinecone, Weaviate, or pgvectorML frameworks: PyTorch, TensorFlow, or JAXYou'll help shape these choices as we build. Why This Role MattersReal-world impact: Your models will help prevent workplace injuries and save livesUnique datasets: Access to proprietary construction data that competitors don't haveGreenfield opportunity: Define the AI strategy from day oneDomain expertise: Work directly with construction industry veteransMission-critical scale: Your models will power verification for facilities where downtime isn't an optionPedigree: A Stanford StartX company (elite sub-1% accelerator)Ready to Build?