QA Engineer, AI Systems for Federal Programs

We hire people
comfortable working
at the edge of what is known.

IOS is a small team building systems for high-stakes environments. Every role touches real federal programs, real research, and real deployment.

How we work

If you want to work on hard problems with clinicians, defense researchers, and engineers who take both the theory and the delivery seriously, we'd like to hear from you.

01

Research and engineering work side by side. People who do well here own their work end to end.

02

Federal compliance is a shared responsibility, not a separate department. Everyone is fluent in what we ship under.

03

A small, technically dense team. The work crosses neuroscience, signal processing, FPGAs, and federal architecture.

6 roles open

01 — Operations

Full-time

Ypsilanti / Detroit metro
· Hybrid

Directs the day-to-day business operations of IOS, including program management coordination, subcontract administration, federal compliance oversight, and scaling operations as the company grows across multiple active contracts.

Responsibilities

• Manage subcontractor relationships and compliance documentation
• Oversee CMMC Level 2 compliance maintenance and SPRS posting updates
• Interface with federal program managers, contracting officers, and agency representatives
• Build operational infrastructure to support company growth from small to mid-size

Full-time. Hybrid (Ypsilanti-based preferred, Detroit metro acceptable).

Build and ship the AI systems that underpin Ypsilon and our research programs. This role owns model architecture, training pipelines, and production inference systems across multimodal AI, active inference implementations, and neuro-symbolic components.

Responsibilities

• Design and implement multimodal AI architectures (speech, language, behavioral signals)
• Develop active inference and Bayesian modeling components
• Contribute to federated learning and privacy-preserving training pipelines
• Collaborate with research partners on model validation and clinical benchmarking
• Own model performance, reliability, and inference latency in production

• Active inference, Bayesian modeling, or predictive processing experience
• Clinical NLP or biomedical language model experience
• Published work in ML, NLP, or computational neuroscience venues

Full-time. Remote with quarterly on-site collaboration in Ypsilanti.

Build and maintain the secure deployment infrastructure that carries IOS's AI systems into federal environments. This role owns the pipeline from code commit to FedRAMP-aligned deployment.

Responsibilities

• Architect and maintain CI/CD pipelines for ATO-ready deployment
• Implement and maintain CMMC Level 2 and NIST SP 800-171 controls
• Build infrastructure-as-code for Azure Government, AWS Gov Cloud, and Oracle Cloud Infrastructure
• Manage secrets, identity, and access control across regulated environments
• Support SBOM generation, vulnerability scanning, and continuous monitoring
• Partner with AI engineering on secure model deployment and federated learning infrastructure

• Experience with ATO authorization packages
• Zero-trust architecture implementation experience

Full-time. Remote with occasional on-site.

Design secure, scalable cloud architectures for federal AI workloads, including GSA-aligned engagements, DARPA program infrastructure, and commercial pilot deployments.

Responsibilities

• Architect FedRAMP-aligned cloud systems across Azure Government, AWS Gov Cloud, and OCI
• Design secure data pipelines for federated learning, de-identified clinical data, and multimodal inference
• Develop reference architectures for repeated use across IOS engagements
• Partner with Dev Sec Ops  on implementation and with AI engineering on model deployment infrastructure

• Certified Azure Solutions Architect Expert, AWS Solutions Architect Professional, or equivalent
• Experience with AI/ML workload architecture (training and inference at scale)

Full-time. Remote with quarterly on-site.

Own quality assurance for AI-driven systems in regulated and…