Generative AI Engineer Role

OpenDataJobs•Washington, DC

About The Position

Generative AI Engineers build production applications around foundation models and large language models. They turn model capabilities into tools for search, drafting, summarization, information extraction, multimodal work, and guided action, with the controls and evidence needed to understand how those tools behave. The role concentrates on application and context engineering. Generative AI Engineers connect models to authoritative knowledge, tools, and workflows; design prompts and structured outputs; evaluate quality, grounding, safety, security, latency, and cost; and monitor the application after release. They treat fluent output as something to test, not proof that the system is correct.

Requirements

A strong application-engineering foundation, including programming, APIs, testing, version control, service integration, and production debugging.
Practical experience with foundation-model integration, prompt and context design, structured outputs, model selection, and failure analysis.
Working knowledge of retrieval systems, embeddings, search, knowledge stores, document ingestion, and data-access controls.
Evaluation discipline across answer quality, retrieval relevance, grounding, safety, security, latency, cost, and user outcomes.
The judgment to constrain tools and agents, design human-review paths, document limitations, and respond when production behavior changes.

Nice To Haves

An application engineer who can work with fast-moving model capabilities without chasing every new release.
Ability to choose architectures by evidence, make uncertainty visible, and separate a convincing demonstration from a dependable service.
Ability to think in complete workflows: sources, context, models, tools, permissions, people, and failure paths.
Experience collaborating with domain experts, data and software engineers, security and privacy specialists, and accountable owners to decide where generation helps and where deterministic methods or human judgment should remain in control.

Responsibilities

Build grounded assistants and knowledge applications with retrieval-augmented generation (RAG), hybrid or vector search, citations, metadata filtering, and source-level access controls.
Develop drafting, summarization, classification, extraction, and transformation services exposed through user interfaces or APIs.
Create agents and multistep workflows with defined tool schemas, constrained permissions, approval gates, memory boundaries, replay, and exception handling.
Design evaluation systems with curated test cases, task-specific rubrics, retrieval measures, grounding checks, safety and security tests, and regression thresholds.
Implement operational pipelines for versioning prompts and configurations, comparing models, tracing execution, monitoring quality and cost, collecting feedback, and responding to incidents.