It was a privilege to speak at GDG Bletchley DevFest 2025 on Friday, November 15th, where I presented “The AI in Your Pocket: Building Offline Gemini Apps with the Browser.” This session explored how Gemini Nano can be leveraged directly within the browser to create a new generation of offline-first, off-device web applications and Chrome Extensions.
The talk focused on empowering users with instant results, ensuring privacy, and maintaining functionality even in environments with limited or no connectivity.
While cloud-based AI offers immense power, it often comes with inherent challenges that can impact user experience. We discussed the critical need for instant results, the hurdles of limited connectivity, paramount privacy concerns when sensitive data leaves the device, and the common barriers of limited usage or paywalls. These factors highlight the need for more robust, on-device solutions.
The solution lies with Gemini Nano, Google’s most efficient AI model, now integrated directly into the Chrome browser. This integration ushers in a new era of offline-first, off-device web applications and Chrome Extensions, prioritizing speed, privacy, and accessibility. The core concept is to run Gemini Nano directly within the user’s browser, bringing powerful AI capabilities to the edge.
The architecture leverages a powerful suite of built-in AI APIs, which are the key enablers for on-device inference. These APIs allow for local processing and are available offline, ensuring data privacy and instant results. We delved into:
During the session, we explored the practical aspects of implementing these APIs, including how to check for API availability, utilize Chrome flags (chrome://flags, chrome://On-device-internals) for development, and manage the download of AI models with progress monitors.
We then took a deep dive into the specifics of each API:
Prompt API: We demonstrated both non-streaming (await session.prompt()) and streaming (session.promptStreaming()) requests, highlighting the flexibility in interacting with Gemini Nano. A key feature discussed was session cloning (languageModel.clone()), which allows for maintaining independent contexts for different topics, ensuring that the original session remains unchanged. We also touched upon constraining the model’s output to a JSON schema for structured responses.
Writer API: This API allows for content creation with fine-grained control over tone, format, length, expectedInputLanguages, and sharedContext. We showed how writer.write() can be used to generate tailored text.
Proofreader API: This powerful tool offers corrections for grammar, spelling, and punctuation, provides labels for error types, and even gives explanations in plain language, making it invaluable for ensuring clear communication.
Rewriter API: We illustrated its capability to transform text, such as changing the tone of an angry bug report into a professional, positive request. The rewrite() method, combined with options for tone, format, and length, allows for sophisticated text refinement.
Translator & Language Detector APIs: These APIs enable seamless on-device language understanding and translation. We showed how to create a translator with specified source and target languages, and how the LanguageDetector can identify the language of text with a confidence score.
Summarizer API: The summarizer’s capabilities are extensive, including summarizing text, suggesting titles and headings, creating concise summaries, and generating teasers. Options like sharedContext, type, format, and length provide flexibility in how content is condensed.
The session concluded with an exciting look into the future. Gemini Nano in the browser ushers in a new era of AI, empowering developers to create next-generation, high-performance, and privacy-centric AI experiences that work seamlessly whether users are online or offline.
The key benefits of this approach are profound:
Attendees left equipped with practical knowledge on how to implement, test, and debug these powerful built-in AI APIs, ready to build the next generation of intelligent web experiences.
It was a fantastic session, and a huge thank you to GDG Bletchley for hosting!
Event Link: DevFest Bletchley Park 2025
Presentation Slides: The AI in Your Pocket
Speaker: Olorunfemi Davis, Senior Platform Engineer at Inflo
Hashtags: #DevFest #GDGBletchley #Gemini #OfflineAI #EdgeComputing #WebDev #PlatformEngineering
A recap of my session at GDG Ado-Ekiti's 'Road to DevFest 2025', where I presented on using Gemini Nano for high-performance, offline-first browser extensions.
Recap of a GDG Cloud North West session in Manchester, exploring offline Gemini Nano apps in the browser, AI APIs, and practical developer insights.
Exploring AI-powered prompt engineering and visual asset generation using Gemini and Gemini Flash 2.5 Image for the DevFest 2025 Avatar creator.
A recap of the presentation on Building Smarter Apps with Microsoft Copilot Studio and .NET at the Birmingham .NET & Xamarin User Group.
A recap of my talk at the .NET Meetup North East, where we explored the developer's role in an AI-first world, diving into Copilot Studio, .NET 9, and Semantic Kernel.
A recap of my lightning talk at .NET Liverpool, where I explored the new AI-first strategy in .NET 9, the Microsoft.Extensions.AI library, and how Semantic Kernel orchestrates AI intelligence in .NET applications.
Recap of our in-person .NET Sheffield event: Building a cultural StoryBot with MCP, C#, and AI—blending cultures.
Exploring the intersection of technical innovation and cultural storytelling. Recap of a .NET York session on building a cultural StoryBot with MCP, C#, and AI.