Yohan Lasorsa is a principal developer advocate at Microsoft, a Google Developer Expert for Angular, and the creator and lead maintainer of ngx-rocket, an open-source toolkit for building production-ready Angular applications. His work focuses on making modern web development more practical, from application scaffolding to AI-assisted coding, debugging, and browser-native inference.
Earlier in his career, Lasorsa worked as a software engineer and cloud architect. He subsequently built ngx-rocket around recurring Angular development needs, including testing, localization, documentation, and extensible application generation. His other open-source projects include backslide, a Markdown-based presentation tool; smoke, a file-based mock server; and hads, a documentation tool for Node.js.
At Microsoft, Lasorsa has written about adapting Angular applications to foldable and dual-screen devices and helped JavaScript developers apply generative AI. His practical examples include a serverless retrieval-augmented chatbot built with LangChain.js and a serverless MCP agent with tool calling and execution tracing.
- Reusable coding-agent skills: Lasorsa uses repository-specific instructions and modular SKILL.md files to turn coding agents into repeatable development workflows. One workflow retrieves a GitHub issue, implements the requested feature, records a Playwright verification video, and sends a mobile preview for human review before the issue can be closed.
- Browser-native multimodal AI: He has demonstrated browser-based inference that combines an uploaded product image with a text prompt and returns a structured JSON review. Running the model locally avoids sending every interaction to an external inference API, although browser support and interfaces remain experimental.
- Debugging inside the running application: He applies browser-integrated AI to console errors, failed network requests, performance traces, and CSS changes, including workflows that write browser edits back to application source files.
- Agent-readable documentation: He uses Angular’s llms.txt resources to illustrate how current framework documentation can help coding agents avoid outdated APIs and obsolete examples embedded in their training data.
At AI Engineer Europe 2026, Lasorsa and AWS developer advocate Olivier Leplus explored websites built for both human visitors and software agents. Lasorsa emphasized reusable coding workflows, contextual browser debugging, local multimodal inference, and current documentation; Leplus demonstrated emerging WebMCP techniques for exposing website actions to agents.
