Project Athena
Python
An AI-powered voice assistant built on local LLMs with RAG capabilities. Runs on Apple Silicon and NVIDIA Jetson edge hardware with Home Assistant integration, real-time speech processing, and multi-room audio coordination.
AI, cloud, and platform engineering
We build production AI and cloud platforms.
We work directly with the founders and engineering leaders who own the outcome. No hand-off, no account team in between. Engagements run from a two-week strategy sprint to a multi-month build, scoped to what actually needs shipping.
- 5
- public repositories
- 30+
- services running in our lab
- 10+
- published articles
Senior engineering, principal‑led.
AGILE Services Group builds production AI systems, cloud platforms, and the infrastructure under them. We work with startups that need to ship fast, growth-stage SaaS teams scaling past early product/market fit, enterprise teams modernizing legacy stacks, and federal programs where compliance and security can't slip.
The practice is AI-first: generative AI strategy, LLM integration, RAG pipelines, AI agents and multi-agent governance frameworks, and the AWS and Kubernetes platforms that run them in production.
Every engagement is led by the principal and staffed with senior specialists chosen for the work in front of them. Architecture and implementation stay together: the engineers who design your system are the engineers who build it.
Based in Baltimore. Fully remote across the US. On-site engagements in the DMV, Philadelphia, and the broader Mid-Atlantic.
You have direct access to the people doing the work.
- Founded
- 2024
- Based
- Baltimore, MD
- Availability
- Remote (US) · on-site DMV, Philadelphia, Mid-Atlantic
- Compliance frameworks
- HIPAA
- Open source
- 5 public repositories
What we do
Services
Hands-on consulting for teams that need senior engineering talent, not slide decks. AI-first, private-sector first, remote-ready.
Generative AI & LLM Implementation
End-to-end LLM product engineering: prompt design, RAG pipelines, evaluations, and production deployment. Model selection across OpenAI, Anthropic Claude, Gemini, and open-source options served through Ollama or vLLM.
Generative AI & LLM consultingHealthcare Technology
Compliant, secure platforms in healthcare and benefits administration: HL7 FHIR integration, benefits-processing systems, and PHI-handling architecture built in from the start.
Healthcare & FHIR consultingTechnical & AI Due Diligence
Architecture, AI capability, security, and team assessment for investors, acquirers, and boards, including AI claims validation — an independent read on the technology before capital moves.
AI & technical due diligenceAI Agents & Multi-Agent Systems
Production multi-agent architectures on LangGraph and custom frameworks, with role-based governance, pre-flight validation, and automatic rollback. Used for engineering automation, research, customer support, coding agents, and domain-specific workflows. Related open-source work: VisionTest.ai and Mozart Orchestration.
AI Strategy & Fractional AI Leadership
Where to apply generative AI for real business outcomes: use-case identification, build-vs-buy decisions, model selection, evaluation strategy, and 6–12 month execution roadmaps. Also available as fractional CTO, fractional AI leader, or fractional Chief Architect for startups and growth-stage companies needing senior technical leadership without a full-time hire.
Cloud & Kubernetes Architecture
Production-grade infrastructure on AWS, multi-cloud, and hybrid environments. Kubernetes platform design, cost optimization, migrations, and modernization for commercial and regulated workloads. From network topology to storage strategy, systems built to perform under real-world conditions.
DevOps & Platform Engineering
Internal developer platforms and CI/CD pipelines that accelerate delivery. Container orchestration, GitOps, infrastructure as code, observability, and self-service tooling that lets teams move fast safely. SRE practices for teams scaling past early stage.
Security & Compliance
Zero-trust architecture, SSO/OIDC, IAM governance, and compliance frameworks for commercial and regulated workloads. Access governance, identity federation, and the audit-ready logging that regulated environments require.
How work moves
Six stages, three review gates — our own open-source multi-agent orchestration pipeline, Mozart Orchestration.
- Research Source and verify every claim
- Plan Explicit phases, reviewed before code
- Review Parallel specialist review
- Implement Phase by phase, gated diffs
- Validate Tests, accessibility, performance
- Ship Deploy, verify, report back
Work
What we've built
We prove architecture in our own lab before it reaches a client environment, and release some of what we build as open source.
Infrastructure & AI
Lab
A Kubernetes cluster running 30+ services, including SSO, GitOps, media infrastructure, document management, and AI workloads. We use it as a proving ground for architectural decisions before they reach production.
Four-node Proxmox cluster on a 10 GbE network with a 120 Gbps Ceph backend, running Kubernetes and 30+ services.
node-01
Proxmox · 96 GB RAM
node-02
Proxmox · 96 GB RAM
node-03
Proxmox · 96 GB RAM
node-04
Proxmox · 96 GB RAM
384 GB RAM across four nodes
~5 TB Ceph (16 TiB raw)
22 TB NAS
208 GB Apple Silicon
Kubernetes · 30+ services
Open source
5 repositories · 10+ articles
SourceBridge
Go
A codebase field guide and context layer for unfamiliar systems. SourceBridge analyzes repositories to help engineers understand complex codebases fast: architecture, conventions, and critical paths surfaced automatically.
VisionTest.ai
TypeScript
AI-powered visual regression testing platform. Captures screenshots, compares against approved baselines using pixel-level and AI perceptual analysis (SSIM, LPIPS, DINOv2, VLM), and surfaces diffs through an approval workflow. Self-hosted, so your data stays on your network.
AgentPulse
TypeScript
A real-time monitoring dashboard for AI coding agent sessions across Claude Code and Codex CLI — live session state, chat history, and orchestration from one command center.
Mozart Orchestration
Shell
A multi-agent orchestration pipeline for Claude Code: DELIVER, AUDIT, and DIAGNOSE workflows that conduct specialist subagents through research, planning, review, implementation, and verification, with live narration.
$ mozart deliver "add rate limiting to the API"
> research · plan · parallel specialist review · implement
> validate · ship · report back
Leadership
Founder and Principal
Jay Stuart
Founder & Principal Architect
Jay Stuart founded AGILE Services Group in 2024. He has 26 years in production software across cloud governance, distributed systems, security and compliance, and healthcare technology.
Roughly a decade of that work has been federal: NOAA, HHS, FDA, VA, CMS, and IRS. As technical lead on federal RFP and re-compete efforts at Booz Allen Hamilton he contributed to over $50M in awarded contracts (HHS, FDA). As Chief Architect at a healthcare technology company he built multi-agent engineering pipelines with role-based governance, pre-flight validation, and automatic rollback, taking an internal delivery cycle from three months to two weeks.
- Certifications
- AWS Certified Solutions Architect · ITILv3 · CMMI Level 3
- Education
- Albany State University, Computer Science & Mathematics
Selected writing
April 1, 2026
Notes from the work
Notes on applied AI, multi-agent systems, software architecture, and the evolving relationship between engineering teams and the AI tools they rely on.
AI Writes My Code. It Also Breaks It. So I Built This.
Building a self-hosted visual regression platform to catch UI breakage that AI-written code introduces and unit tests miss.
Read on Medium
March 28, 2026
March 28, 2026
March 16, 2026
March 12, 2026
March 12, 2026
I Replaced Alexa With a Fully Local AI Assistant — Here's What Actually Happened
How a weekend experiment became a 30-service local AI assistant with no cloud dependencies.
AIWe're Shipping Code We Don't Understand — And Calling It Progress
AI isn't just helping us write code. It's changing our relationship to understanding it. And the gap between what we ship and what we comprehend is growing.
SoftwareMob Programming, Eight Years Later — And Why AI Might Make It Even Better
A retrospective on introducing mob programming, what made it work, and why AI tooling might revive the practice.
TeamsAI Agents Have Write Access to My Code — Here's How I Sleep at Night
A four-layer recoverability design for AI agents with write access: Forgejo mirrors, Harbor caching, immutable S3 backups, dependency scanning.
DevOpsThe AI Speed Trap — And How I'm Building My Way Out
A six-phase engineering loop for keeping AI coding speed without losing architectural discipline.
Process
Technology
Tools we work with
Production-tested tools and platforms, including legacy entries we still support.
Infrastructure
- AWS (EKS, ECS, RDS, S3)
- Kubernetes
- CloudFormation
- Terraform
- Karpenter
- VPC / Networking
- CloudWatch
Security & Compliance
- HIPAA
- IAM / SSO
- Active Directory
- FIPS 140-2
Languages & Frameworks
- Python
- Java
- Go
- JavaScript / TypeScript
- PHP
- Perl
- C# / .NET
- Bash
- C++
Data & Storage
- PostgreSQL
- MySQL
- Oracle
- SQL Server
- Redis
- DB2
- RDS
DevOps & Delivery
- ArgoCD / GitOps
- CI/CD Pipelines
- Docker
- Helm
- Ansible
- GitHub Actions
- MuleSoft
AI & ML
- LLMs / GPT
- RAG Pipelines
- LangGraph / LangChain
- AI Agents
- Multi-Agent Systems
- Self-Healing QA
- Vector Databases
- Ollama
Questions
Frequently asked
Does AGILE Services Group offer AI consulting?
Yes — AI consulting is a primary practice: generative AI strategy, LLM integration, RAG pipelines, AI agents and multi-agent systems, and AI platform and infrastructure work. Engagements are scoped from a two-week strategy sprint to a multi-month production build-out, sized to what actually needs shipping.
Does AGILE work with private-sector companies, or only federal clients?
Both. The practice serves startups, growth-stage SaaS, and enterprise teams, and carries a federal delivery record since 2014 — NOAA, HHS, FDA, VA, CMS, and IRS. Federal work is one credential among several, not the whole practice.
Where is AGILE Services Group based, and what regions do you serve?
Based in Baltimore, Maryland. The firm works fully remote across the United States, with on-site engagements across the Mid-Atlantic corridor: Baltimore, Washington, D.C., Northern Virginia (Arlington, Alexandria, Tysons, Reston), Bethesda, Annapolis, and Philadelphia.
What AI models, platforms, and tools do you work with?
Commercial and open-source models, from OpenAI, Anthropic Claude, and Google Gemini to self-hosted options deployed on AWS or Kubernetes. Project Athena, one of the firm's open-source repositories, runs local models on Apple Silicon and NVIDIA Jetson edge hardware.
Can AGILE be hired as a fractional CTO, AI leader, or Chief Architect?
Yes. Fractional CTO, fractional AI leader, and fractional Chief Architect engagements run remote or on-site — a fit for startups and growth-stage companies that need senior technical and AI leadership without a full-time hire.
What kinds of AI projects does AGILE take on?
Generative AI strategy, LLM-powered product features, RAG systems over proprietary data, AI agent and multi-agent workflows, AI platform and infrastructure design, and technical or AI due diligence for investors.
What non-AI consulting does AGILE offer?
Cloud architecture (AWS, multi-cloud), Kubernetes platform design and migrations, DevOps and platform engineering, CI/CD, observability, SRE, and healthcare technology, including HIPAA and HL7 FHIR integration.
How do I get in touch to start a consulting conversation?
Email info@agileservicesgrp.com to start a conversation, use Calendly to book a 30-minute call, fill out the contact form on this page, or reach out on LinkedIn directly.
Book time
Schedule a conversation
30 minutes, no sales pitch. Tell us about the AI product you're building, the platform you're scaling, or the problem you're trying to solve. Remote or on-site (Baltimore, DC, Philly, NoVA).
Get in touch
Let's build something
Whether you're shipping your first AI product, scaling a platform past early stage, evaluating a vendor's AI claims, or modernizing a regulated workload, tell us about your project. Private sector first. Federal clients welcome. Fully remote across the US; on-site in Baltimore, DC, Philadelphia, and Northern Virginia.
Or email us directly at info@agileservicesgrp.com • LinkedIn • GitHub