AI Engineer- Virtual

Alight Solutions•Deerfield, IL

1d•$120,000 - $220,000•Remote

About The Position

We are building AI that meets them in those moments, agentic assistants that answer real questions and carry out transactions on a participant’s behalf, and it runs in production today. The platform this team builds is what makes those experiences fast, reliable, and safe at scale. This is hands-on engineering with a direct line to millions of users, and we are hiring builders at every level to take it further. We build the AI engineering platform that powers AI delivery across Alight. We own the developer lifecycle, deployment infrastructure, observability, quality automation, document ingestion pipelines, and reusable agent patterns, so delivery teams solve business problems instead of reinventing infrastructure.

Requirements

Python. proficiency (3.10+). This is the primary language across the team.
AWS. Working knowledge of core services (IAM, ECS/Fargate, networking) and comfort in cloud-native environments.
Software engineering discipline. Version control (Git/GitHub), code review, testing, documentation, and clean code.
Security awareness. You build things that handle sensitive data, and you think about secrets, permissions, and data protection by default.
Expertise in one of the six: AI agent development (LangGraph or similar, multi-agent orchestration, memory, tool use, guardrails, prompt and workflow design), Data and retrieval engineering (Chunking, embeddings and vector search, RAG, hybrid search, reranking, retrieval evaluation), Quality and evaluation (Agent eval frameworks (DeepEval, Ragas), CI/CD quality gates, groundedness and safety metrics, test automation for AI systems), Platform and infrastructure (CI/CD, Terraform, container orchestration on AWS (ECS, Fargate), secrets and IAM, progressive deployment), Observability and operations (OpenTelemetry tracing, dashboards and alerting, LLM cost and token tracking, feedback loops), Developer experience (MCP servers, devcontainers, CLI and SDK tooling, prompt libraries, the docs and reference implementations others build on).

Responsibilities

The agent patterns, retrieval, and guardrails behind the agentic assistants that answer participant questions and carry out transactions on their behalf
An MCP server that serves a skills registry to AI coding agents right in the IDE
Reusable agent patterns in LangGraph: multi-agent orchestration, memory, tool-use safety, and guardrails
An agent evaluation platform (DeepEval, Ragas) with quality gates that block a bad deployment in CI/CD
Document ingestion and retrieval pipelines: chunking, embedding, vector search, reranking, and citations
CI/CD and deployment automation for AI agents on AWS (ECS, Fargate, Agent Core)
AI-enabled devcontainers with tooling, validation, and testing built in
OpenTelemetry instrumentation, dashboards, and LLM cost and token tracking for production workloads
Governance controls turned into pipeline checks, with evidence collected automatically
Semantic caching, rate limiting, and graceful degradation for agent workloads under load
Reference implementations, SDK and CLI tooling, and the docs other engineers build on