Chief Scientist, AI/ML

• The model layer. Project 137’s shared AI/ML stack: model selection and routing across frontier and open-weight models, fine-tuning on mission data, retrieval, and inference at IL5/IL6 and the tactical edge. Every pillar consumes it.
• The harness. The AI/MLOps and agent harness that takes a model or agent from experiment to cleared production: pipelines, evaluation, deployment, monitoring, and governance, running on our own on-prem compute.
• Mission evals. The test-and-evaluation framework that proves a model or agent works on a real mission task, fails safely, and holds up under adversarial pressure. It is our answer to “how do you know it works?”, You lead a core team of 2-3 senior scientists and engineers. You also hold technical authority over the AI/ML engineers embedded in each pillar: you set the methods, the stack, and the hiring bar, and the pillars execute on them.
• The AI/ML vision. Define and socialize Project 137’s AI/ML roadmap across all four pillars. Decide where we lead, where we follow, and where we stop spending.
• LLMs on mission data. Fine-tune, distill, and serve models on scarce, sensitive, and classified data. Own retrieval, model routing, and inference optimization from the cloud to disconnected and SWaP-constrained environments.
• Computer vision and ATR. Lead automatic target recognition and sensor exploitation: detection, classification, and tracking across EO/IR, SAR, RF, and emerging sensor modalities, and fusion into a decision-grade picture with our Decision Advantage team.
• Agentic AI and evals. Architect multi-agent systems that do real mission work, with the guardrails, auditability, and evaluation that make them accreditable. Build the mission eval harness every pillar uses before anything ships.
• AI/MLOps and the harness. Own the path from notebook to cleared production: training and eval pipelines, model registry, deployment, monitoring, drift detection, and rollback. Make it repeatable so pillars ship without reinventing it.
• On-prem compute. Shape and run the AI hardware in our lab: GPU clusters, scheduling, storage, and networking, plus air-gapped and classified deployment patterns. Decide what runs on-prem, what runs in government cloud, and what runs at the edge.
• IP that Quantum Sky keeps. Quantum Sky retains IP by default. You decide what we build, partner for, or buy, and you shape the wrappers we build around frontier and open-weight models.
• The technical face of AI/ML. Brief program-of-record owners and CIO/CTO/CDAO offices across DoW, IC, and FedCiv. Speak at conferences, publish, and represent Quantum Sky in the AI/ML community. Support technical diligence on AI/ML M&A targets.
• The bench. Recruit, develop, and keep AI/ML talent across Project 137. You are a reason strong people join., * AI/MLOps and harness development. Built the pipelines, eval harnesses, and model lifecycle tooling that take models and agents from experiment to production, including monitoring, drift detection, and rollback.
• On-prem hardware. Designed, stood up, or operated on-prem AI compute: GPU clusters, high-speed interconnect, workload scheduling, and storage, ideally in air-gapped or classified settings.
• Mission context. Delivered AI/ML to DoW, IC, or Federal Civilian customers, or in another regulated, high-consequence environment. Comfortable with accreditation, IL5/IL6, and Responsible AI requirements.

Required:

• The track record. 12+ years in AI/ML, including 5+ years leading AI/ML teams. You have fielded AI systems in production that people depend on, not only prototypes and papers.
• LLMs in production. Hands-on fine-tuning, retrieval, evaluation, and inference optimization of large language models on domain-specific data.
• Computer vision. Shipped computer vision or ATR systems: detection, classification, tracking, or sensor fusion on real-world imagery or signals.
• Agentic systems. Recent hands-on work building multi-agent systems, with the evaluation and guardrails that make them trustworthy. You know the difference between an agent demo and an agent you would put in front of a mission., * Research pedigree. PhD in computer science, electrical engineering, statistics, or a related field. Peer-reviewed publications at top venues, patents, or widely used open-source work.
• A recognized voice. Conference talks, public writing, or community leadership that people in AI/ML know you for.
• Active TS/SCI.
• Edge AI. Fielded models in disconnected, intermittent, or SWaP-constrained environments.
• Adversarial ML. Hands-on experience red-teaming models and agents, or defending them against attack.
• Quantum-curious. Interest in where quantum and quantum-inspired methods may help AI/ML, and the judgment to tell a real result from hype.

Clearance:

• Active Secret or TS/SCI preferred; strong candidates may be sponsored.