F
Senior AI Engineer
Floor & Decor Holdings
3 hours ago
Full-time
On-site
Atlanta, Georgia, United States
Senior AI Engineer
The Senior AI Engineer will design, build, and operate the framework and platform capabilities that enable agentic solutions across Floor & Decor — and will build flagship agentic products on that platform to prove it out. The role covers the full delivery lifecycle: planning, designing, configuring, testing, implementing, documenting, and maintaining AI solutions deployed on Microsoft Azure and integrated with the company's operational systems. This is a hands-on senior individual contributor role. This is not a research or data science position. We are looking for an engineer with first-principles command of backend systems who has moved into AI engineering — someone who treats LLMs and agents as components in a well-architected distributed system, and holds them to the same standards of reliability, observability, security, and cost control as any other production dependency. Technical leadership here is exercised through architecture, code quality, and influence rather than through direct reports. Minimum Eligibility Requirements: Engineering Foundation 7–10 years of professional software engineering experience, with increasing scope and ownership. Deep proficiency in backend engineering in at least one of C#/.NET, Java, Node.js/TypeScript, or Python. We care about first-principles understanding of backend systems, not a specific language — the ability to pick up a new stack quickly matters more than which one you arrived with. Python is a plus, not a requirement. Service-based and microservice architectures, RESTful API design, and asynchronous service communication — including API versioning, contract design, error semantics, and backward compatibility. Cloud-based serverless microservices in production (Azure Functions, Container Apps, or equivalent). Event-driven architecture: queues, pub/sub, idempotency, retries, and dead-letter handling. Solid data fundamentals: relational and NoSQL data modeling, query performance, and transactional correctness. Demonstrated ownership of code quality — automated testing, code review, and CI/CD as normal practice, not overhead. AI Engineering Designed and deployed agentic solutions — shipped and operated in production, not prototypes or POCs. We are interested in depth of agent delivery experience; retrieval-augmented generation is one tool in that kit, not the definition of the job. Harness engineering — practical command of the scaffolding that surrounds a model and determines whether an agent actually works in production: context construction and management, tool and function-call interfaces, multi-step planning and control flow, memory and state, structured output, guardrails and validation, human-in-the-loop checkpoints, and graceful failure and fallback behavior. Azure OpenAI services — LLM APIs, Azure AI Search (vector/index), and associated Azure infrastructure. Azure AI Foundry or AWS Bedrock experience also applies. Agent and LLM orchestration frameworks (LangChain, LangGraph, or similar). MCP (Model Context Protocol) or other tool-calling/function-calling patterns for LLM-to-system integrations. AI-specific failure and risk modes — hallucination, prompt injection, data leakage, non-determinism, runaway tool loops — and concrete techniques for mitigating them. Sound judgment on when an LLM or agent is and is not the right tool, and how to bound and validate its output. AI-Augmented Development Daily working fluency with agentic coding tools (Claude Code, Cursor, GitHub Copilot, Devin, or equivalent). Claude Code is our standard. A credible, specific point of view on where these tools accelerate delivery, where they do not, and how to review and test model-generated code responsibly. Able to speak to measurable impact on their own or their team's throughput and quality. Cloud, Security, and Delivery Cloud environments (Azure preferred) including compute, storage, networking, and IAM fundamentals. Cloud security fundamentals: identity and access management, secrets management, and network boundaries. Agile/scrum methodologies — sprint ceremonies, story estimation, backlog grooming. Angular or comparable modern front-end frameworks — working knowledge. Clear written and verbal communication with both technical and non-technical partners; comfortable with ambiguity and able to move from a vague business problem to a scoped, specified, shippable increment. Education Bachelor's degree in Computer Science, Engineering, or a related field — or equivalent professional experience. Demonstrated capability is weighted above credentials. Essential Job Functions: Agentic Platform & Framework Engineering Design, build, and evolve the shared framework that lets teams across the organization build, deploy, and operate agentic solutions — a paved road covering agent scaffolding, tool and integration interfaces, context and memory management, evaluation, guardrails, and observability. Turn one-off agent implementations into reusable building blocks: service templates, harness components, MCP server patterns, evaluation harnesses, and deployment pipelines. Design and maintain MCP integrations with enterprise systems — inventory, merchandising, and the systems that follow — so any agent built on the platform can be grounded in live operational data without rebuilding the plumbing. Define and document the standards, contracts, and reference architectures that agentic workloads at Floor & Decor are built against. Make and document architectural decisions for reliability, scalability, security, and observability of AI services deployed in Azure. Agentic Solution Delivery Design, build, and operate production agentic solutions end to end — from problem framing through deployment and ongoing operation — using them both to deliver business value and to harden the underlying platform. Apply the right technique to the problem: tool calling, multi-step planning, retrieval, memory, human-in-the-loop review, or a deterministic service where an agent is the wrong answer. Build and tune retrieval where retrieval is warranted — chunking strategies, vector indexing, retrieval ranking, and context engineering on Azure AI Search. Contribute across the stack, including an Angular front end and a Python-based service and LLMOps layer — picking up front-end work to get a feature over the line rather than handing it off. Support the transition of mature agentic products to partner teams: documentation, runbooks, and knowledge transfer that let a solution outlive its original builders. Evaluation, Observability & Operations Establish evaluation and regression testing as a first-class part of the platform — eval sets, LLM-as-judge scoring, task-level success metrics, and regression gates in CI — so changes to prompts, models, tools, or retrieval ship with evidence rather than intuition. Own evaluation pipelines for retrieval-based components using Ragas, tracking faithfulness, answer relevance, and context precision across releases. Instrument agentic systems for observability with Langfuse alongside Azure Monitor — tracing, latency, token and cost attribution, tool-call success rates, quality signals, and failure modes — and act on what the telemetry shows. Manage AI cost and performance: token budgeting, caching, model routing and right-sizing, and latency optimization. Own production services — on-call participation, incident response, and post-incident follow-through. Troubleshoot and resolve complex issues in agent behavior, retrieval quality, hallucination, latency, and integration reliability. Engineering Practice & Collaboration Apply spec-driven development: turn ambiguous business asks into clear specifications and acceptance criteria before code, and keep specs and implementation in sync. Set and model the standard for AI-augmented development on the team: effective use of agentic coding tools, plus the review discipline that has to come with it. Conduct code reviews and mentor peers, fostering a culture of quality and continuous learning — through technical influence rather than direct reporting lines. Partner with product managers and business stakeholders to identify workflows worth automating, translate them into technical requirements, and help prioritize the backlog. Act as a technical consultant to other teams adopting the platform — helping them build well on it rather than around it. Participate in agile ceremonies — sprint planning, retrospectives, and daily stand-ups — as a senior voice on the team. Communicate technical trade-offs and architectural decisions clearly to both technical and non-technical audiences. Partner with security, data, and platform teams on data governance, PII handling, prompt injection defense, and responsible use. Innovation & Quality Evaluate emerging agent capabilities, tooling
The Senior AI Engineer will design, build, and operate the framework and platform capabilities that enable agentic solutions across Floor & Decor — and will build flagship agentic products on that platform to prove it out. The role covers the full delivery lifecycle: planning, designing, configuring, testing, implementing, documenting, and maintaining AI solutions deployed on Microsoft Azure and integrated with the company's operational systems. This is a hands-on senior individual contributor role. This is not a research or data science position. We are looking for an engineer with first-principles command of backend systems who has moved into AI engineering — someone who treats LLMs and agents as components in a well-architected distributed system, and holds them to the same standards of reliability, observability, security, and cost control as any other production dependency. Technical leadership here is exercised through architecture, code quality, and influence rather than through direct reports. Minimum Eligibility Requirements: Engineering Foundation 7–10 years of professional software engineering experience, with increasing scope and ownership. Deep proficiency in backend engineering in at least one of C#/.NET, Java, Node.js/TypeScript, or Python. We care about first-principles understanding of backend systems, not a specific language — the ability to pick up a new stack quickly matters more than which one you arrived with. Python is a plus, not a requirement. Service-based and microservice architectures, RESTful API design, and asynchronous service communication — including API versioning, contract design, error semantics, and backward compatibility. Cloud-based serverless microservices in production (Azure Functions, Container Apps, or equivalent). Event-driven architecture: queues, pub/sub, idempotency, retries, and dead-letter handling. Solid data fundamentals: relational and NoSQL data modeling, query performance, and transactional correctness. Demonstrated ownership of code quality — automated testing, code review, and CI/CD as normal practice, not overhead. AI Engineering Designed and deployed agentic solutions — shipped and operated in production, not prototypes or POCs. We are interested in depth of agent delivery experience; retrieval-augmented generation is one tool in that kit, not the definition of the job. Harness engineering — practical command of the scaffolding that surrounds a model and determines whether an agent actually works in production: context construction and management, tool and function-call interfaces, multi-step planning and control flow, memory and state, structured output, guardrails and validation, human-in-the-loop checkpoints, and graceful failure and fallback behavior. Azure OpenAI services — LLM APIs, Azure AI Search (vector/index), and associated Azure infrastructure. Azure AI Foundry or AWS Bedrock experience also applies. Agent and LLM orchestration frameworks (LangChain, LangGraph, or similar). MCP (Model Context Protocol) or other tool-calling/function-calling patterns for LLM-to-system integrations. AI-specific failure and risk modes — hallucination, prompt injection, data leakage, non-determinism, runaway tool loops — and concrete techniques for mitigating them. Sound judgment on when an LLM or agent is and is not the right tool, and how to bound and validate its output. AI-Augmented Development Daily working fluency with agentic coding tools (Claude Code, Cursor, GitHub Copilot, Devin, or equivalent). Claude Code is our standard. A credible, specific point of view on where these tools accelerate delivery, where they do not, and how to review and test model-generated code responsibly. Able to speak to measurable impact on their own or their team's throughput and quality. Cloud, Security, and Delivery Cloud environments (Azure preferred) including compute, storage, networking, and IAM fundamentals. Cloud security fundamentals: identity and access management, secrets management, and network boundaries. Agile/scrum methodologies — sprint ceremonies, story estimation, backlog grooming. Angular or comparable modern front-end frameworks — working knowledge. Clear written and verbal communication with both technical and non-technical partners; comfortable with ambiguity and able to move from a vague business problem to a scoped, specified, shippable increment. Education Bachelor's degree in Computer Science, Engineering, or a related field — or equivalent professional experience. Demonstrated capability is weighted above credentials. Essential Job Functions: Agentic Platform & Framework Engineering Design, build, and evolve the shared framework that lets teams across the organization build, deploy, and operate agentic solutions — a paved road covering agent scaffolding, tool and integration interfaces, context and memory management, evaluation, guardrails, and observability. Turn one-off agent implementations into reusable building blocks: service templates, harness components, MCP server patterns, evaluation harnesses, and deployment pipelines. Design and maintain MCP integrations with enterprise systems — inventory, merchandising, and the systems that follow — so any agent built on the platform can be grounded in live operational data without rebuilding the plumbing. Define and document the standards, contracts, and reference architectures that agentic workloads at Floor & Decor are built against. Make and document architectural decisions for reliability, scalability, security, and observability of AI services deployed in Azure. Agentic Solution Delivery Design, build, and operate production agentic solutions end to end — from problem framing through deployment and ongoing operation — using them both to deliver business value and to harden the underlying platform. Apply the right technique to the problem: tool calling, multi-step planning, retrieval, memory, human-in-the-loop review, or a deterministic service where an agent is the wrong answer. Build and tune retrieval where retrieval is warranted — chunking strategies, vector indexing, retrieval ranking, and context engineering on Azure AI Search. Contribute across the stack, including an Angular front end and a Python-based service and LLMOps layer — picking up front-end work to get a feature over the line rather than handing it off. Support the transition of mature agentic products to partner teams: documentation, runbooks, and knowledge transfer that let a solution outlive its original builders. Evaluation, Observability & Operations Establish evaluation and regression testing as a first-class part of the platform — eval sets, LLM-as-judge scoring, task-level success metrics, and regression gates in CI — so changes to prompts, models, tools, or retrieval ship with evidence rather than intuition. Own evaluation pipelines for retrieval-based components using Ragas, tracking faithfulness, answer relevance, and context precision across releases. Instrument agentic systems for observability with Langfuse alongside Azure Monitor — tracing, latency, token and cost attribution, tool-call success rates, quality signals, and failure modes — and act on what the telemetry shows. Manage AI cost and performance: token budgeting, caching, model routing and right-sizing, and latency optimization. Own production services — on-call participation, incident response, and post-incident follow-through. Troubleshoot and resolve complex issues in agent behavior, retrieval quality, hallucination, latency, and integration reliability. Engineering Practice & Collaboration Apply spec-driven development: turn ambiguous business asks into clear specifications and acceptance criteria before code, and keep specs and implementation in sync. Set and model the standard for AI-augmented development on the team: effective use of agentic coding tools, plus the review discipline that has to come with it. Conduct code reviews and mentor peers, fostering a culture of quality and continuous learning — through technical influence rather than direct reporting lines. Partner with product managers and business stakeholders to identify workflows worth automating, translate them into technical requirements, and help prioritize the backlog. Act as a technical consultant to other teams adopting the platform — helping them build well on it rather than around it. Participate in agile ceremonies — sprint planning, retrospectives, and daily stand-ups — as a senior voice on the team. Communicate technical trade-offs and architectural decisions clearly to both technical and non-technical audiences. Partner with security, data, and platform teams on data governance, PII handling, prompt injection defense, and responsible use. Innovation & Quality Evaluate emerging agent capabilities, tooling