Jeff Kazzee

[](https://jeffkazzee.dev)

Software engineer focused on agentic operations, local-first tools, and practical AI systems

I build durable software around messy work: internal tools, typed knowledge systems, local-first products, and agent workflows that keep state, expose decisions, and leave proof behind.

I use AI heavily, but I treat it as engineering leverage: scope the work, build the artifact, verify it, document it, and make the next run easier.

Start Here

| Link | What to look for |
| --- | --- |
| Portfolio | Public project writeups, live surfaces, and the broader story of what I build. |
| Vivary | My main systems project: typed memory, workspace state, review gates, graph search, and agent coordination. |
| Obscura | Local-first Electron product work: AI media workflows, cost visibility, SQLite, packaging, and release surfaces. |
| Open World Factbook | Data/product execution: an open country and territory database with a public web surface. |
| Writing | Practical notes on building with AI, agent workflows, and turning messy work into durable systems. |
| Contact | Open to software engineering, AI-enabled internal tools, agent operations systems, and docs-heavy product work. |

What I Build

• Agentic operations systems: structured memory, visible state, reusable skills, human approval gates, and verification loops.
• Local-first products: desktop and web apps that keep files, costs, prompts, outputs, and history inspectable.
• Typed knowledge systems: Markdown knowledge bases with schema checks, graph structure, source/current-truth separation, and CI-friendly validation.
• Technical docs and resource sites: practical guides, searchable field resources, beginner onboarding, and proof-first writing.
• Delivery systems: planning, testing, implementation, review, learning loops, release hygiene, and repo cleanup.

Selected Work

| Project | Proof of capability |
| --- | --- |
| Vivary | Agent-workspace framework. Ships Python and npm surfaces for typed memory, project state, reusable skills, review gates, graph search, coordination, docs, CI, and release tracking. |
| Obscura | Local-first Electron media studio for AI creative work: chat, prompts, model/cost visibility, image workflows, gallery lineage, local SQLite, and alpha release packaging. |
| Open World Factbook | Open-source country and territory database covering 262 countries and territories, with a public web surface at worldfactbook.xyz. |
| sdlc-skills | Installable agent skills that encode the software-development lifecycle: plan, design, test, implement, review, and learn. Built for disciplined AI-assisted engineering across agent runtimes. |
| Throughline | Tiny Agent OS for keeping AI work visible, grounded, cheap to load, and human-led through a simple Ask -> retrieve -> act -> verify -> learn -> gate loop. |
| Braincheck | Zero-dependency Python typechecker for Markdown knowledge bases. Validates YAML frontmatter schemas, supports safe autofix, JSON output, and CI-friendly exit codes. |
| Loam | Folder-as-type knowledge layer for Markdown. Derives metadata from path, git, and headings so frontmatter carries only irreducible signal. |
| Minimal Zo Brain | Specification-first setup prompt that interviews a user and creates a right-sized AI-readable second brain around real work, relationships, sources, tasks, and approval boundaries. |
| AI Starter Skill Pack | Beginner-safe agent skills for AI onboarding, prompt debugging, learning paths, project ideas, and honest career proof. |
| Career Coach | Evidence-first career operating system with agent skill instructions, Obsidian-friendly templates, scorecard tooling, and tests. |
| Dreaming / Flywheel | Background pattern-finding and workspace maintenance skills for Zo, Claude Code, Codex, and local multi-agent workflows. |
| jeffkazzee.dev | Astro portfolio and writing surface for public projects, proof artifacts, and Substack mirrors. |

What I Bring To A Team

• I can take an ambiguous product idea and turn it into a scoped build plan, working artifact, documentation, and verification path.
• I understand agent workflows as software systems: context boundaries, instructions, skills, memory, source truth, review gates, and failure recovery.
• I write for humans and agents at the same time: READMEs, specs, runbooks, prompts, release docs, onboarding flows, and durable project state.
• I care about inspectability: local files, visible costs, provenance, plain Markdown, testable contracts, and small tools people can actually understand.
• I ship across the stack: TypeScript, Python, Astro, React, Next.js, Electron, SQLite, YAML, Markdown/MDX, GitHub Actions, Vercel, and Cloudflare.

Work I Am Looking For

I am most useful in roles or scoped projects involving:

• AI-enabled product prototyping and internal tools
• Agentic workflow and operations design
• Developer-experience and documentation-heavy engineering
• Local-first desktop or web apps
• Knowledge-base, second-brain, or typed-document systems
• Public technical content backed by working software
• Repo rescue: README, docs, tests, release surfaces, and credibility polish

I am self-taught, fast-moving, and proof-first. The throughline across my work is simple: build the thing, verify the thing, document the thing, and make the next session easier than the last one.

Links

• Portfolio: jeffkazzee.dev
• GitHub: github.com/Jeff-Kazzee
• Substack: jeffkazzee.substack.com
• X: @JeffKazzee
• Bluesky: @jeffkazzee.bsky.social

Collapse