AI systems builder · MCP server developer · product-minded automation engineer · Long Beach, CA
I build practical AI tools that survive contact with real workflows: Model Context Protocol servers, agent-facing CLIs, repo-native product surfaces, launch docs, evaluation loops, and domain software for people doing messy creative or operational work. My current work lives primarily at Kyanite Labs, where I build open-source MCP servers and AI tools for video editing, localization, estimation, creative coding, and domain-specific software.
My background is a little unusual in a useful way: 11 years in enterprise operations, learning systems, training, enablement, and HR technology, plus hands-on open-source engineering across Python, TypeScript, automation, media tooling, and AI agent infrastructure. I care about the parts where demos usually break: install paths, docs, metadata, tests, local execution, permissions, and whether a human can actually understand what the system is doing.
| Project | What it is | Why it matters |
|---|---|---|
| KyaniteLabs/mcp-video | Video editing MCP server, Python library, and CLI with FFmpeg and Hyperframes tooling. | Lets Claude Code, Cursor, Codex-style agents, and other MCP clients inspect and edit video locally. |
| KyaniteLabs/DialectOS | Spanish dialect localization MCP server and CLI. | Helps teams translate and QA Spanish across regional variants, register, docs, markdown, and product copy. Published as @espanol/mcp. |
| KyaniteLabs/openglaze | Ceramic glaze calculator, UMF analyzer, CTE estimator, and recipe manager. | Brings self-hosted open-source software to potters and ceramic artists. |
| KyaniteLabs/Epoch | Time, cost, schedule, token, and risk estimation tools for AI agents. | Gives planning agents a grounded way to reason about delivery risk, and gets better from real estimated-vs-actual usage feedback. |
| KyaniteLabs/dev-learning-archaeologist | Git-history learning diagnostic and codebase archaeology tool. | Turns repository history into evidence-backed learning plans and engineering narratives. |
Epoch is the project I most want people to try right now. AI agents are terrible at estimating time unless they have tools, reference classes, and calibration feedback. Epoch gives them those tools.
Install it in Claude Code, Cursor, VS Code, Windsurf, or any MCP client:
claude mcp add epoch -- npx @puenteworks/epochThen use it on real tasks and share anonymized feedback: estimated duration, actual duration, task complexity, team context, model used, token/cost estimate, and whether the risk forecast was right. That data is how Epoch gets sharper.
- MCP servers and agent tools for Claude Code, Cursor, Codex-style workflows, and local automation.
- Python and TypeScript CLIs, APIs, test suites, release pipelines, and package metadata.
- AI-search-friendly documentation: README,
llms.txt,ai.txt, examples, changelogs, docs sites, and launch copy. - Workflow automation for media, localization, creative coding, learning systems, and operational teams.
- QA systems that look for silent failures, stale metadata, missing docs, broken install paths, and unverified claims.
I like tools that are accesible, understandable, local-first when privacy matters, boring where reliability matters, and polished where public trust matters. The best AI systems are not just clever prompts. They are inspectable workflows with clear contracts, honest docs, good defaults, and enough tests to catch the weird edge cases before users do.
- Kyanite Labs: AI tools, MCP servers, domain software, and repo-native productization.
- PuenteWorks: AI consulting and workflow automation for small businesses.
- Capital Group: talent development, Workday Learning, Microsoft 365 Copilot adoption, learning operations.
- Southern California Edison: project/program analysis, SAP SuccessFactors, operations support.
- PIMCO and American Red Cross: training operations, enablement, and bilingual instruction.
model context protocol mcp server ai agents agent tools python typescript ffmpeg automation video editing automation local-first ai developer tools ai evaluation repo metadata llms.txt spanish localization creative coding open-source software
- Want the clearest public proof? Start with mcp-video, DialectOS, Epoch, or OpenGlaze.
- Want the company/project surface? Start with Kyanite Labs or kyanitelabs.tech.
- Want to understand my operating style? Look for the docs, tests, release notes, examples, repo metadata, and launch checklists. That is usually where the real quality signal is.
Open to AI systems, developer tools, product engineering, solutions engineering, and operations automation roles where the job is to make useful systems real.