AI Engineer

Job Overview

As a Staff AI Engineer, you define and drive the architecture of AI and agentic systems across multiple teams and product domains. This is a senior individual‑contributor leadership role: you influence high‑impact architectural decisions, evolve the practices and standards for building agentic AI, and turn experimental AI capabilities into reliable production systems. You set direction for multi‑agent orchestration, production RAG (hybrid search, re‑ranking, and query routing), tool and MCP integration, and the evaluation and observability stack that keeps them dependable.

You mentor senior engineers and represent AI engineering in cross‑functional and strategic initiatives. A background in classical ML is an asset; the primary requirement is a proven track record of shipping production agentic AI.

• Define and drive technical direction for AI and agentic systems, and contribute to the AI platform roadmap across teams
• Influence architecture decisions for compute, cloud, and AI infrastructure across teams
• Lead the design of large‑scale AI/LLM systems: inference platforms, APIs, and distributed architectures
• Architect production multi‑agent systems end‑to‑end: orchestration, state management, tool integration, and failure handling
• Define and drive best practices and standards for AI/LLM systems across teams (agent design, evaluation, observability, reliability)
• Lead complex production debugging and incident response across teams, and harden the resulting fixes into platform guardrails
• Mentor senior engineers and emerging technical leaders, raising the engineering bar
• Lead technical design reviews and architecture decision records (ADRs) for critical AI infrastructure
• Contribute to capacity planning and cost optimization strategies for AI/LLM infrastructure

• Define and drive vector database and RAG architecture decisions across systems and teams: structured RAG, hybrid search (dense + sparse + keyword), re‑ranking, and query routing
• Lead multi‑agent platform architecture decisions: runtime selection, orchestration patterns, and enterprise integration strategy
• Set the technical direction for MCP (Model Context Protocol) adoption and agent runtime infrastructure
• Shape agent infrastructure adoption: evaluate and standardize frameworks, tooling, and deployment patterns for agentic AI
• Architect evaluation infrastructure for non‑deterministic LLM systems: synthetic golden‑set generation, hierarchical weighted scoring (component, composite, and system‑level F1), bootstrap confidence intervals, and paired A/B comparison, treating a change as real only when it is both statistically significant and clears a minimum effect size
• Gate deployments on eval results: tiered regression thresholds (hard‑gate vs monitor components) wired into CI so a measurable quality regression blocks the release, with observability via tracing across multi‑step chains and tool calls and drift detection on LLM inputs and outputs
• Drive LLM cost optimization at scale: model routing, caching, batching, token budget management, and provider cost analysis

• Master’s or Ph.D. degree in Computer Science, Software Engineering, or a related field
• 10+ years of professional experience in ML/AI or software engineering, including 4+ years in senior or staff‑level roles with production system ownership
• Demonstrated engineering leadership, including driving technical strategy and influencing cross‑team decisions
• Expertise in platform and distributed‑systems architecture at scale: model serving, APIs, data platforms, and AI/LLM infrastructure
• Hands‑on experience architecting and operating production agentic AI or LLM systems (multi‑agent workflows, production RAG, tool and MCP integration)
• Deep understanding of embedding models, retrieval algorithms, and vector database internals
• Strong production debugging, reliability, and incident‑response skills
• Experience building rigorous evaluation for non‑deterministic AI systems, including statistical methods (such as bootstrap confidence intervals and minimum effect‑size thresholds) to separate genuine quality changes from…