Staff Agentic AI / Data Engineer

About the Role

AMD's Applied AI team works with the world's most demanding AI operators — frontier labs, NeoCloud providers, and AI-native companies — to make AMD Instinct GPU infrastructure the easiest place to build and run AI. As an Agentic Data Engineer, you will build the data and agent systems that sit at the heart of this mission: production agentic AI applications running on AMD clusters, the data pipelines and memory/context databases that give those agents durable knowledge, and the skills frameworks and evaluation infrastructure that make agent behavior reliable, measurable, and safe.

Your work spans two surfaces. Externally, you build agentic systems and their data foundations on customer AMD deployments — the reference implementations customers adopt when they move from inference to agents. Internally, you build the Applied AI team's own intelligence layer: engagement memory databases, fleet and telemetry data pipelines, and agent-executable skills libraries that encode deployment knowledge so every customer engagement makes the next one faster.

This is a production engineering role. The systems you build run live, get depended on, and are held to production standards for quality, provenance, and security.

Responsibilities

• Build production agentic AI systems on AMD Instinct GPU infrastructure: agent orchestration, tool/function calling (including MCP-based integrations), skills frameworks, and streaming inference integration against ROCm-based serving stacks (vLLM, SGLang)
• Design and operate the memory and context data layer for agentic applications: vector, graph, and relational stores, embedding pipelines, retrieval and context-engineering strategies, and the freshness, provenance, and access-control policies that govern them
• Build the Applied AI team's engagement memory and fleet data infrastructure: pipelines that ingest deployment telemetry, incident histories, and field knowledge into structured, queryable, agent-consumable form
• Develop and maintain the skills library: reusable, versioned, agent-executable encodings of deployment and operational expertise, with the testing and review gates required before agents or engineers rely on them
• Build evaluation infrastructure for agentic systems: regression suites, LLM-as-judge pipelines, behavioral test harnesses, and production quality monitoring
• Harden agentic systems against real-world failure modes, including prompt injection through retrieved context and memory stores, data poisoning, and tool-misuse paths
• Create the reference architectures and open artifacts that make AMD the credible platform for agentic workloads, contributing upstream to the open-source agent, serving, and data ecosystem
• Partner with customer-facing engineers on live engagements: your systems deploy into customer environments, and you support their production behavior

Requirements

• Deep software engineering experience with significant production data engineering: pipelines, storage systems, and data quality at scale (level flexible for exceptional candidates)
• Hands-on experience building LLM-powered and agentic applications in production: agent frameworks and orchestration, RAG and context engineering, tool calling, and multi-step workflows
• Depth in at least one memory/context storage paradigm — vector databases, graph databases, or hybrid retrieval architectures — and informed opinions about when each is wrong
• Experience designing evaluation frameworks for non-deterministic systems
• Strong Python; working fluency with modern data stack tooling (orchestration, streaming, warehouse/lakehouse) and containerized deployment on Kubernetes
• Familiarity with GPU inference serving (vLLM, SGLang, or comparable) and the performance characteristics of LLM workloads; ROCm/AMD Instinct experience a strong plus
• Security-conscious engineering instincts, particularly around untrusted content flowing into model context
• Open-source contribution history in the AI/ML or data infrastructure ecosystem is a plus

Qualifications

Bachelor's or Master's degree in Computer Science, Computer Engineering, Data Engineering, or equivalent practical experience.

This role is not eligible for visa sponsorship.

Benefits

Benefits offered are described at AMD benefits at a glance.