Olivier Leplus is a senior developer advocate at Amazon Web Services and a Google Developer Expert in web technologies focused on how artificial intelligence is reshaping browsers, web applications, and software development. His work treats the web as both a platform for running AI locally and an interface that autonomous agents must learn to navigate.
Earlier in his career, Leplus worked on conversational technology at Botfuel, where he described multilingual chatbot middleware that detects users’ languages, translates messages, and maintains conversational context. He subsequently worked in developer relations at Microsoft before joining AWS, where his work expanded from feature-flag deployment strategies to Amazon Bedrock agents and agent-driven web applications.
With Florent Lacroute and Frederic Visticot, he co-created the Amazon Bedrock Web Tools sample, an open-source browser-automation assistant combining Playwright, visual understanding, form interaction, human checkpoints, and the Model Context Protocol.
- Agent-ready web applications: Leplus advocates giving software agents explicit, structured access to website functionality. In his AI Engineer Europe session with Yohan Lasorsa, he demonstrated experimental WebMCP tools that expose actions such as adding products to a shopping cart and convert existing HTML forms into machine-readable interfaces. He emphasizes that these emerging APIs remain experimental and can change quickly.
- Browser-native AI as progressive enhancement: His guide to built-in browser AI describes running summarization, translation, and other AI features on users’ devices, with hosted models available when local capabilities are missing. The approach balances privacy and operating costs against browser compatibility, hardware requirements, and model downloads.
- Observable, tool-using agents: Leplus uses Chrome DevTools and the Model Context Protocol to let agents inspect network requests, capture screenshots, and compare application performance across connection speeds. His introduction to Strands Agents frames agent development around models, tools, instructions, debugging, and production deployment.
- Cloudagotchi: His open-source virtual pet connects an ESP32 device to AWS services, combining cloud-managed state, serverless processing, and AI-generated spoken news. The project makes distributed cloud architecture tangible through connected-device security, hardware constraints, and a deliberately playful interface.