Build vertical agents. Own end-to-end agents for high-value workflows across research, software, hardware, data, and operations—from problem definition through production use.
Establish the agent foundation. Build reusable primitives for models, tools, orchestration, context, memory, retrieval, permissions, human approval, and long-running execution.
Create the context layer. Connect agents to trusted data through APIs, pipelines, MCP servers, and integrations with clear provenance and access control.
Own evaluation and reliability. Build benchmarks, regression tests, tracing, monitoring, and failure-analysis loops across quality, latency, cost, security, and resilience.
Advance frontier agent usage. Evaluate new models, coding agents, SDKs, and patterns, then turn useful capabilities into maintainable systems rather than demos.
Support strategic initiatives. Help company leadership apply agents and analytical systems to market and customer intelligence, partnerships, fundraising, diligence, scenario analysis, and executive decisions.
Provide technical leadership. Set architecture and engineering standards, review designs and code, and create reusable patterns for the technical team.
Qualification
5+ years in software engineeringExperience at a frontier AI labExperience building evaluationsStrong product judgmentBonus QualificationsExperience with post-trainingExperience with MCP infrastructureBackground in robotics
Required
5+ years in software engineering, applied AI, ML systems, or a related field, with strong zero-to-one technical judgment.
Experience at a frontier AI lab, leading AI company, or comparable team working at the edge of current model capabilities.
A track record shipping production agentic systems that real users depend on—not only prompts, prototypes, or demos.
Deep experience with frontier model APIs, tool use, orchestration, context engineering, retrieval, memory, and multi-step workflows.
Experience building evaluations, regression tests, observability, and production failure-analysis loops for AI systems.
Exceptional fluency with AI-native development workflows using Claude Code, Codex, Cursor, agent SDKs, or equivalent systems, with a rigorous understanding of where agents work and fail.
Strong product judgment: able to turn an ambiguous decision or workflow into a useful, secure, measurable system.
Requires 5 days/week in-office collaboration with the team.
Bonus Qualifications
Experience with post-training, model evaluation, inference, or research infrastructure at a frontier lab or model company.
Experience with MCP infrastructure, developer platforms, knowledge graphs, RAG, or secure enterprise integrations.
Background in robotics, autonomous systems, industrial automation, or another technically complex physical-world domain.
