Projects

Selected projects showing how I build, test, and reason about AI developer systems. AI Systems Lab is an experimental learning environment; Agent Skills is the more directly reusable developer workflow project.

EXPERIMENTAL

AI Systems Lab

An experimental learning and evaluation environment for AI model systems, exploring provider-neutral routing, model lifecycle, compatibility, benchmarks, and local llama.cpp and cloud-hosted OpenAI-compatible capability boundaries.

Technologies

Pythonllama.cppOpenAI-compatible APIsHugging FaceJSONL

Outcomes

  • ✓Workload-specific model aliases across local llama.cpp and cloud-hosted OpenAI-compatible provider backends
  • ✓Compatibility gates for visible completions and structured tool calls
  • ✓Layered throughput, server-path, and quality evaluation workflows
Read case studySource

Agent Skills

An open-source collection of reusable, evidence-backed agent skills that turn practical engineering judgment into dependable software-development workflows.

Technologies

Agent SkillsMarkdownYAMLGitHub

Outcomes

  • ✓Evidence-based GitHub public-readiness audits
  • ✓Workflow guidance, evaluation criteria, and reporting structure kept separate
  • ✓MIT-licensed foundation for future agent tooling
Read case studySource

erikfryscok.com

A public engineering home base delivered as a fast, static site with a practical workflow for planning, previews, owned infrastructure, and domain email.

Technologies

AstroTypeScriptTailwind CSSCloudflare PagesGitHubProton Mail

Outcomes

  • ✓Static production site served without a server-side runtime
  • ✓Cloudflare-managed domain and DNS with Proton Mail domain email
  • ✓Public GitHub Issues and project board for delivery planning
Read case studySource

More selected work will appear here as it is ready to share publicly and can support useful, evidence-based lessons.