AI Agent Engineer

Lexsi Labs is a frontier AI lab focused on building aligned, interpretable, and safe superintelligent systems . Our work spans alignment methodologies, interpretability-led system design, and foundational model research across structured, tabular, and enterprise data. We build AI systems intended for real-world deployment, where transparency, auditability, and robustness are first-class constraints. We operate with a flat structure, high autonomy, and a strong bias toward engineers who take full ownership of what they build, from architecture to production behavior. The Role We are building a high-autonomy AI Engineering Agent that can execute end-to-end AI engineering workflows on top of the Lexsi backend. The system is designed to reason over complex objectives, plan multi-step actions, interact with tools and data systems, evaluate outcomes, and iterate over long horizons while remaining interpretable and aligned. As a Sr AI Agent Engineer , you will be responsible for defining and building the core agent architecture. You will make foundational decisions about how the agent reasons, plans and executes actions, maintains memory and state, and has its decisions inspected and explained. This role requires deep hands-on engineering and the ability to operate in ambiguous environments with ownership from day one. What You’ll Build

• End-to-end AI Engineering Agents that can autonomously perform tasks such as experiment design, evaluation, analysis, and reporting, rather than merely assisting a human in these workflows.
• Agent architectures that combine reasoning, planning, tool orchestration, and memory , with explicit handling of long-horizon execution, partial failures, and self-correction.
• Deep integrations with the Lexsi backend , including internal evaluation systems, data pipelines, and alignment tooling, as well as enterprise APIs and proprietary data sources.
• Agents designed for enterprise and regulated environments , where every action, decision, and output must be inspectable, explainable, and defensible. Responsibilities
• Own the architecture and implementation of the AI Engineering Agent , making early design decisions around planners, reasoning loops, memory models, and execution control that will shape the system long-term.
• Build robust tool orchestration and execution layers , ensuring the agent can interact reliably with internal services, external APIs, and data systems, and recover gracefully from failures.
• Design and implement memory and state management , enabling agents to operate across long-running tasks, retain relevant context, and reason over past actions and outcomes.
• Embed alignment, safety, and interpretability into the system design , working closely with research teams to ensure that agent behavior is auditable and controllable by construction.
• Stress-test agent behavior in real-world conditions , identifying edge cases, failure modes, and distribution shifts, and iterating on the system to improve reliability and correctness.
• Set a high engineering bar , contributing to design discussions, code reviews, and technical decision-making with a strong sense of ownership and accountability. Requirements
• Significant experience building and shipping complex AI or ML-heavy systems , where you have owned architecture and seen systems evolve in production, not just prototypes.
• Hands-on experience with agentic AI systems and frameworks , such as ReAct-style agents, AutoGPT-like systems, LangChain, LangGraph, Semantic Kernel, or similar, with a clear understanding of their limitations and failure modes.
• Strong backend engineering fundamentals , including advanced Python proficiency, experience building APIs and services, and familiarity with databases, data pipelines, and cloud infrastructure.
• A systems-thinking mindset , with the ability to reason about performance, reliability, cost, safety, and interpretability as interconnected design constraints.
• Comfort working in ambiguous, fast-moving environments , where problems are loosely specified and ownership is expected rather than assigned. Strong Bonus Signals
• Experience building long-horizon or stateful agent systems that must reason over time rather than single-turn interactions.
• Prior exposure to AI alignment, interpretability, or safety tooling , especially in production or enterprise settings.
• Experience working in regulated or high-stakes domains , where system behavior must be explained and audited after deployment.
• A track record of debugging and improving systems that behaved unpredictably in the real world. What This Role Is Not
• Not a junior or early-career role.
• Not a prompt-engineering or “glue code” position.
• Not a research-only role detached from production systems. If you need tightly scoped tasks, detailed step-by-step instructions, or long onboarding periods, this role will not be a good fit. Interview Process
• A short intro call to vibe check - not more than 15 minutes with the CEO.
• A technical deep-dive round centered on system design and failure analysis with Tech and AI leads
• A final conversation for alignment and mutual fit. We move quickly and expect candidates to do the same. We value substance over polish and execution over rhetoric.