Localizeflow
Founder / AI infrastructure / 44M+ words processed
A GitHub-native localization workflow service that watches documentation repositories, processes translation workloads outside fragile CI paths, and opens reviewable pull requests.
Work that connects model behavior to software delivery.
My work sits at the boundary between AI systems and developer infrastructure: localization jobs, pull request workflows, RAG systems, Azure AI tutorials, and the operational layer that keeps automation reviewable.
Selected work
Co-op Translator
Microsoft OSS maintainer / 55 languages
An open-source CLI and Python API for translating Markdown, notebooks, and image text while preserving structure, links, frontmatter, and maintainer review loops.
Microsoft PhiCookBook
Contributor / developer education / 40 closed PRs
Tutorials and workflow improvements around Phi fine-tuning, Prompt Flow, Azure AI Studio, Azure AI Foundry, and Responsible AI evaluation.
Azure AI writing and workshops
Technical educator / 120K+ views
Practical writing and sessions on Azure AI Search, Azure OpenAI, RAG, fine-tuning, safety, and production LLM workflows.
Impact
44M+ words processed
AI localization workloads across production repositories.
835+ automated PRs
GitHub-native localization pull requests across Microsoft OSS.
55 languages
Technical learning content kept synchronized for global learners.
120K+ public views
Azure AI and LLM engineering writing read by developers.
Operating style
Production evidence over demo polish
I care about queues, retries, timeouts, review loops, formatting drift, observability, and the small operational details that decide whether AI work keeps working.
Automation with human control
The best AI systems make output easier to review, retry, and trust. They do not hide uncertainty from the people responsible for shipping.
Open systems that teach
Writing, talks, and open-source work are part of the system. Good infrastructure becomes more useful when its tradeoffs are visible.