Agentic AI Engineer

Trilagen•McLean, VA

22h

About The Position

We build AI agents that actually work in enterprise environments — not prototypes, not demos. We need engineers who can own the entire agent stack: a production frontend, a robust backend, a properly secured API and identity layer, a memory architecture that scales, and LLM integrations that are model-agnostic and built to last. You'll be deployed on client engagements as the lead technical architect and builder of agentic systems running in AWS, OCI, and Azure. You'll work directly with client stakeholders, translate complex requirements into working systems, and leave behind infrastructure clients can operate and extend. You'll also help Trilagen productize our delivery approach as we scale the practice. If you've only ever built agents that run on your laptop, this isn't the role. If you've shipped agentic systems into production cloud environments and know exactly what breaks and why — we want to talk.

Requirements

4+ years of software engineering experience
At least 2 years building and shipping LLM-powered or agentic applications in production cloud environments
Hands-on depth with at least two of the three major LLM providers: Anthropic Claude, OpenAI, and Google Gemini — at the API level
Full-stack proficiency: Python (FastAPI, Flask, or similar) backend, React or Next.js frontend, REST and WebSocket API design
Production experience on at least two of: AWS, Azure, OCI — with real deployments
Demonstrated ability to design and implement agent memory and retrieval systems using vector databases and RAG
Strong command of AI security practices: auth, RBAC, secrets management, audit logging, and prompt-level safeguards

Nice To Haves

Experience with all three LLM providers (Anthropic, OpenAI, Gemini) and model-agnostic orchestration patterns
Okta and/or SailPoint ISC integration experience
Cloud certifications: AWS Solutions Architect, Azure Solutions Architect, OCI Architect

Responsibilities

Full-stack agent development: Design and build the entire application, including a React or Next.js frontend with streaming, real-time agent UX; a Python or Node.js backend that orchestrates agent logic, manages state, and exposes clean APIs; and containerized, cloud-deployed services. Own the repo, CI/CD pipeline, deployment, and runbook.
Multi-cloud deployment: Deploy production agent systems on AWS (Lambda, ECS/Fargate, Bedrock, API Gateway), Oracle OCI (OKE, Functions, AI Services), and Azure (AKS, Azure OpenAI Service, Azure Functions). Understand the tradeoffs between platforms and advise clients on deployment strategies.
LLM integration and model strategy: Architect model-agnostic integration layers for Anthropic Claude, OpenAI, and Google Gemini. Select, swap, and benchmark models for specific agent tasks.
Agentic architecture: Design systems for multi-step work, including tool use, ReAct and plan-and-execute loops, agent-to-agent orchestration, human-in-the-loop checkpoints, retry and failure recovery, and cost/latency optimization. Utilize frameworks like LangGraph, CrewAI, AutoGen, and the Model Context Protocol (MCP).
Memory and context layer: Design and implement memory architectures for production agents, including short-term conversational context, long-term persistent memory, RAG pipelines with vector databases (Pinecone, pgvector, OpenSearch, Weaviate), semantic search, and hybrid retrieval strategies.
Security and identity layer: Build security envelopes around agent systems, including OAuth 2.0 / OIDC authentication, API key lifecycle management, role-based access control, secrets management (AWS Secrets Manager, Azure Key Vault, OCI Vault), audit logging, prompt injection defense, and data residency controls.
Consulting: Run discovery sessions, write technical design documents, manage client expectations, and own delivery end to end.