Senior AI Engineer

Overview

Systems Planning and Analysis, Inc. (SPA) delivers high‑impact technical solutions to complex national security issues. With over 50 years of business expertise and consistent growth, SPA is known for continuous innovation for our government customers in the United States and abroad.

In this role, you will lead the development of systems built on foundation models ranging from small language models to large language and multimodal models. You will ensure these capabilities are deployed securely across local, hybrid, and cloud model backends—including Microsoft Azure (including GCC High and Azure Government), Google Cloud Platform, and on‑premises/edge infrastructure.

This is a bridge role: your primary depth is in AI engineering and delivery, with strong working fluency in cloud security and Dev Sec Ops  practices. You will partner with Infrastructure and Security teams to deliver secure, mission‑aligned AI at scale in highly regulated environments.

• Build and deploy production AI applications using Azure AI Foundry, Azure OpenAI Service, and Copilot Studio, accounting for service availability differences between Azure Commercial, Azure Government, and GCC High environments.
• Select and right‑size models for mission requirements, balancing capability, cost, latency, and deployment constraints across small, medium, and large foundation models (e.g., SLMs such as Phi, frontier LLMs, embedding and multimodal models).
• Engineer agentic AI systems, including multi‑agent frameworks (e.g., Semantic Kernel, Lang Graph, Auto Gen) and tool‑use pipelines, including Model Context Protocol (MCP)‑based integrations.
• Develop RAG architectures using Azure AI Search and vector stores, including embedding pipelines, document chunking strategies, and grounding‑data governance (Purview/DLP integration).
• Orchestrate model endpoints and optimize inference workloads across local, hybrid, and remote backends—including managed cloud endpoints, self‑hosted inference on AKS, and local/on‑prem serving runtimes (e.g., ONNX Runtime, vLLM, Foundry Local).
• Design backend‑agnostic application architectures with abstraction layers that allow models to be swapped or routed between local, hybrid, and cloud endpoints based on data sensitivity, latency, cost, and connectivity constraints.
• Implement MLOps/LLMOps practices: model evaluation harnesses, AI red‑teaming, prompt versioning, and telemetry/observability for AI applications.

• Ensure AI workloads conform to GCC High and Azure Government constraints, including CUI handling, data residency, customer‑managed key requirements, and appropriate placement of inference (local vs. cloud) based on data classification.
• Support secure multi‑cloud operations across Azure and GCP, partnering with Infrastructure teams.
• Configure AI security guardrails, content safety controls, DLP policies, gateway policies, and alignment safeguards, informed by the NIST AI Risk Management Framework and OWASP Top 10 for LLM Applications.

• Implement AI traffic governance and secure inspection using modern AI gateways.
• Maintain secure inter‑cloud connectivity and workload visibility using NSGs, firewall rules, traffic mirroring/visibility tooling, and service‑to‑service authentication (OAuth 2.0, Entra managed identities, workload identity federation).
• Embed automated security validation (SAST/DAST) into CI/CD pipelines.

• U.S. citizenship.
• Bachelor's degree in computer science, data science, cybersecurity, IT, or a related field.
• 5–7 years in enterprise software or systems engineering, with a strong recent focus on cloud‑scale AI architectures.
• 3–5 years building AI/ML solutions, including 1–2 years hands‑on with Azure OpenAI, Azure AI Foundry, Copilot Studio, or equivalent foundation‑model platforms.
• Experience working across model scales and deployment models—small/specialized through large foundation models—deployed via managed cloud endpoints, self‑hosted, or local runtimes, and selecting appropriately for the use case.
• Experience developing agentic AI systems and integrating…