Owns the compute infrastructure that trains and serves AI models — GPU cluster management, distributed training pipelines, and inference serving at scale. A natural pivot for DevOps/backend infra engineers.

An emerging specialization focused on designing how information is structured, retrieved, and fed into an LLMs context window — optimizing for relevance, token efficiency, and reasoning quality rather than model weights themselves.

Builds and maintains the evaluation frameworks that determine whether an AI system is good enough to ship — designing test sets, scoring rubrics, automated graders, and regression suites. A role most product companies currently can't hire fast enough for.

The MLOps discipline adapted for large language models — owns the infrastructure for deploying, monitoring, versioning, and scaling LLM-based systems in production, including cost management and reliability.

Specializes in building autonomous or semi-autonomous AI agents that plan, use tools, and take multi-step actions with minimal human intervention. One of the highest-hiring-volume and most technically cutting-edge roles in 2026.

The broad, catch-all title for engineers building applications on top of foundation models — spanning RAG, agents, fine-tuning, and infrastructure. Overtook "ML Engineer" as the fastest-growing tech title in 2026. Scope varies widely by company size.

Focuses on the technical implementation and integration side of enterprise AI rollouts — configuring, customizing, and troubleshooting a vendor's AI platform within a customer's infrastructure. More platform-configuration focused than the build-from-scratch FDE role.

Builds AI features directly into a company's core product — think in-product copilots, recommendation systems, or automation features. Less client-embedded than an FDE; works within a product engineering org, partnering with product managers and designers.

A more explicitly AI-focused variant of the FDE role, emphasizing model integration and applied ML over general software delivery. Focuses on adapting foundation models, RAG systems, and agentic workflows to a specific customer's data and use case, often at frontier AI labs.

Embeds directly inside a customer's environment (on-site, remote, or in the customer's own cloud) to scope, build, and ship an AI system end-to-end. Owns the outcome — not just the code — acting as a hybrid of software engineer, solutions architect, and customer success manager. Integrates AI (often LLM/RAG-based) into the client's existing data pipelines and workflows, builds domain-specific evaluation frameworks, and iterates with the client until the system is production-grade and trusted.