Generative AI Applications Engineer (Agents & RAG)

Accenture Federal Services•Washington, DC

1d•$103,200 - $203,400

About The Position

At Accenture Federal Services (AFS), we are dedicated to strengthening the US federal government and improving the lives of its citizens through technology and innovation. Our team of over 13,000 professionals works across defense, national security, public safety, civilian, and military health organizations. AFS, a technology company within the global Accenture network, is recognized as a Glassdoor Top 100 Best Place to Work. We foster a collaborative and inclusive community that empowers individuals to grow, learn, and thrive through hands-on experience, certifications, and industry training. We are seeking individuals to join us in driving positive, lasting change that advances critical missions and the government forward. We specialize in building production-ready Generative AI applications for confidential federal programs, prioritizing reliability, privacy, and safety. Our approach focuses on rapid deployment, shipping in weeks rather than quarters, and measuring success through key performance indicators such as latency, reliability, safety, and cost. Confidentiality is paramount, and program details are not publicly disclosed. Specifics will be shared during the interview process for qualified candidates.

Requirements

Built and deployed a production GenAI application (chatbot, copilot, assistant, or enterprise AI).
Hands-on with LLMs GPT, Claude, Llama, Gemini, Mistral, via APIs or self-hosted.
Designed and implemented RAG solutions using embeddings, vector databases, and semantic search for enterprise or mission data.
Shipped AI applications with frameworks like LangChain, LangGraph, LlamaIndex, Semantic Kernel, DSPy, or similar.
Strong Python development for building and integrating AI/ML applications.
Owned AI solutions through full production life cycle, deployment and operational support.
Active U.S. citizenship.

Nice To Haves

Integration with leading cloud AI services or on-prem inference stacks
Background in LLM evaluation, prompt authoring/testing, A/B experimentation, and LLM Ops.
Responsible AI expertise (privacy, security, bias, transparency, human in the loop) and data governance.
Experience implementing tools using agents for API integration and external data access.
Containerization & orchestration (Docker, Kubernetes, VMware) and scripting/automation (Linux Bash, PowerShell).
Prior work in regulated/secure environments (e.g., ATO, STIGs, Zero Trust) with fast shipping.
Familiarity with NVIDIA AI Foundations, OpenAI ChatGPT, and AI assisted dev tools (Cursor, Windsurf, Claude).
Contributions to internal frameworks or opensource; mentorship of engineers.
Clear communication with engineers, PMs, and security/compliance stakeholders.

Responsibilities

Design & ship mission grade GenAI: Build agentic workflows and RAG systems tailored to mission data and environments; target low hallucination, tight p95 latency, and predictable cost.
Agent frameworks & orchestration: Apply patterns from LangChain/LlamaIndex/Semantic Kernel; design task decomposition, tool use, guardrails, and recovery/fallback strategies.
Platform integration (no model training): Implement with AWS Bedrock, Azure OpenAI, Google Vertex AI, Amazon Kendra, and managed services (e.g., Document AI, Gemini, Gemma).
LLM selection & evaluation: Compare models for quality, safety, latency, cost; author/test prompts & policies; deploy with observability and safe rollback/fallback.
RAG done right: Build retrieval pipelines & vector search (Pinecone, Weaviate, OpenSearch, pgvector, FAISS/Chroma); handle data prep, chunking, metadata, and IR-style evals (e.g., NDCG) to maximize signal to noise.
Production rigor: Instrument metrics/logs/traces; run A/B experiments; maintain incident playbooks; and implement safety & compliance guardrails.
SRE & FinOps for AI: Define SLIs/SLOs (quality/latency/safety/cost), run on call and postmortems, reduce MTTR; meter usage and optimize token/spend.
Reusable platform components: Ship SDKs, CI/CD templates, Terraform/IaC modules, evaluation harnesses that accelerate multiple mission teams, not one-off projects.
Operate in real-world constraints: Deliver into hybrid, restricted, or air-gapped environments with Zero Trust principles and audit-ready controls.