GenAI Engineering Lead

Job DescriptionWe're looking for a GenAI Engineering Lead to design and build the agentic systems behind our AI platform on AWS, and to lead the design work that gets them shipped.This is a builder's role first. You'll design multi-agent architectures, write the code for the hard parts yourself, own the technical direction of what we build, and set the patterns the rest of the team follows.Agentic systems change fast. A large part of this job is staying ahead of that — evaluating new frameworks, models, and patterns before we need them, running small experiments to separate what's real from what's hype, and making calls today that won't trap us in six months. You'll bring that research back as working prototypes and clear recommendations.You'll work closely with business stakeholders to turn ambiguous problems into agent designs that hold up in production, and with our GenAI platform and DevOps teams to get them deployed safely.Reports to the Head of GenAI Development. Partners with the Lead GenAI Architect, the Process Automation team, and the Platform, Security and QA teams.ResponsibilitiesDesign and build agentic systems end to end — orchestration, tool and API integration, memory, retrieval, and human-in-the-loop patterns — and stay hands-on in the codebase.Lead a team of developers: set technical direction, review designs and code, and mentor engineers.Own the agent framework stack (LangGraph, Strands, Bedrock AgentCore, and what comes next) and the research behind it — evaluate what's emerging, prototype what looks promising, and make calls today that still hold up in six months.Design the context layer our agents reason over — knowledge graphs, vector stores, and retrieval strategy — owning the data modeling, schema, and embedding decisions behind it.Design and own the caching layer — semantic, prompt, and context caching with invalidation rules that keep results trustworthy — and drive latency and cost through model routing and right-sizing.Own quality and safety: eval harnesses, regression suites, and observability for non-deterministic systems, plus guardrails, tool permissioning, PII handling, and human approval paths for high-impact actions.Partner with business stakeholders to turn real problems into agent designs, work with platform and DevOps to deploy them securely, and leave behind the docs and reference patterns other teams build from.Qualifications10+ years in software engineering, including 2+ building LLM-powered systems, including time as technical lead for a small team — setting direction, reviewing designs and code, and still writing code yourself.Strong Python and modern engineering practice: testing, CI/CD, code review, observability, and iterative delivery.Production experience with LLM-powered systems — breaking domain tasks down into reusable agents, tools, and skills, and shipping agentic workflows and RAG pipelines to real users rather than demos.Depth in the retrieval and context layer: embeddings (including fine-tuned or domain-adapted models), vector search, chunking, and the data modeling and enrichment that determine retrieval quality — plus working knowledge of graph or other structured context stores and of caching strategies for latency and cost.Hands-on with at least one agent framework or runtime (LangGraph, Strands, Bedrock AgentCore, or similar), with MCP or an equivalent tool-integration protocol, and a solid AWS foundation — Lambda, API Gateway, IAM, and Bedrock or equivalent model serving.Experience defining evaluation methodology for non-deterministic systems — rubric-based or LLM-as-judge evaluation, metrics that reflect real task quality, and regression testing that catches drift.A research habit: able to evaluate an emerging model, framework, or pattern quickly, form a defensible opinion, and explain it to stakeholders.Preferred QualificationsRunning vector or graph stores in production — sizing, index tuning, re-embedding and reindexing, backups, and access control — and building GraphRAG or hybrid retrieval on top of them.Tuning a semantic cache in production — setting similarity thresholds, measuring hit rate against false-hit rate, and handling invalidation as underlying data changes.AWS Bedrock, AgentCore, or a comparable enterprise GenAI platform.Fine-tuning, distilling, or otherwise adapting models for specific tasks, and serving self-hosted or in-house models in production.Work in a specialized domain where correctness matters — legal, financial, or regulatory — including delivery in security-sensitive environments.Understanding of MCP servers and gateways that expose internal systems and APIs, with authentication, scoped permissions, versioning, and a vetted catalog so teams reuse tools instead of rebuilding them.The base salary range for this position is $200,000 - $240,000 per year. This range reflects the minimum and maximum base salary we reasonably expect to pay for this role. In addition, this position may be eligible to participate in the relevant business unit’s incentive compensation plan, and other compensation programs as applicable. Eligible employees may participate in a 401(k) program with a generous profit-sharing contribution, medical, prescription dental, and vision coverage; life insurance; disability coverage; paid holidays; vacation; and sick time, subject to plan terms and Company policies.About Bessemer TrustBessemer Trust is a family office, overseeing more than $250 billion in assets for over 3,000 individuals and families of substantial wealth. Its more than 1,300 employees are singularly focused on private wealth management — disciplined investment management, sophisticated wealth planning, comprehensive family office services, and highly personalized client service.Established in 1907 as the family office for Annie and Henry Phipps, Bessemer Trust is in its seventh generation of ownership by the Phipps family. As a self-made entrepreneur, Henry Phipps was a founding partner and chief financial officer of Carnegie Steel.Bessemer Trust retains its original focus as a privately owned and independent wealth manager deeply committed to its mission of providing peace of mind to its clients. Bessemer’s adherence to putting clients’ interests first, fiduciary mindset, and highly collaborative culture are at the heart of everything the firm does.Key Facts:For 119 years, Bessemer Trust has operated continuously in a single line of business, independently owned by one family.Headquartered in New York’s Rockefeller Center, Bessemer Trust has 22 offices in total. Woodbridge, NJ, is one of the firm’s largest offices, which hosts a wide range of technology and operations professionals. In addition to its sizable presence in New York and Woodbridge, the firm provides client service through offices in Atlanta, Boston, Chicago, Dallas, Delaware, Denver, Garden City, Grand Cayman, Greenwich, Houston, Los Angeles, Miami, Naples, Nevada, Palm Beach, San Diego, San Francisco, Seattle, Stuart, and Washington, D.C.To watch a video about Bessemer Trust’s history, click here.To learn more about Bessemer Trust, click here.About Our Employee Rewards and Benefits:We provide exceptional rewards and benefits that are among the best in the industry, giving our people access to a wide range of options, including:Competitive base salary plus discretionary annual bonus for select positionsA 401(k) plan with a generous annual profit-sharing contributionPersonalized development and career opportunities, including tuition reimbursement supportComprehensive medical, dental, and vision plans with zero contributions for employee coverageEmployee assistance (EAP) and wellness programsHybrid work environment: 60% in office, 40% remote for most positionsPaid time off and paid parental leaveEmployer-paid life insurance and short- and long-term disability coverageLegal services and financial wellness plans at no cost to employees Bessemer Trust is committed to creating a diverse and inclusive environment and is proud to be an equal opportunity employer. We encourage candidates of diverse backgrounds to apply.Job DetailsJob Type: Full-timeCategory: Information TechnologySalaried: Salaried