Staff Software Engineer (AI Foundation Model)

About this role

• Technically strategize on Foundation model selection, refactor, design, build, and maintain Merlin’s core AI models,training and inference infrastructure, including distributed training pipelines, experiment tracking, and model registry systems
• Define and drive standards and benchmarks for model evaluation, benchmarking, and regression testing to ensure AI systems meet safety and performance thresholds before deployment
• Architect and own the foundational data pipeline that ingests, processes, labels, and versions flight data for use across all autonomy and AI teams
• Identify and resolve systemic technical bottlenecks that slow down AI development velocity across Merlin’s engineering organization
• Collaborate with AI infra, Simulation, and Flight Software teams to define interfaces and shared abstractions that make the broader stack more coherent and maintainable
• Mentor, build and technically guide junior and mid-level engineers on the AI Foundation team, conducting design reviews and raising the overall quality of the team’s output
• Lead technical scoping and estimation for large AI foundational model projects, breaking ambiguous requirements into actionable engineering plans
• Evaluate and adopt relevant open-source tooling, frameworks, and research, contributing back where appropriate
• Document architecture decisions, system designs, and operational runbooks to a standard that supports safety review and long-term maintainability
• Contribute to hiring by conducting technical interviews and helping define the engineering bar for the AI Foundation team

### Benefits

• Equity in Merlin
• Flexible paid time off
• Comprehesive healthcare
• Parental leave
• Lifestyle spending account
• 401K program- You are a senior individual contributor with deep expertise in building the systems that make large-scale machine learning possible in production
• You’ve experience in the selection of optimum foundation models, adopt/tune for target application needs, designed...