Shao Hang He
Shao Hang He is a serial entrepreneur and senior software developer. He has more than 11 years of experiences in software development. He worked for big companies such as Ubisoft, BlackBerry and Corus Entertainment. He also co-founded several startups (MailMagic.ai, DevFortress, Blossom.team, Geek-it and Fanstories) and raised money from Venture Capital.
Proposals - Montreal 2025
How to build a Google Workspace Add-on
Google Workspace Add-ons can extend Gmail, Calendar, Drive, Docs, Chat and Meet. Unlike Chrome extensions, which are limited to desktop browsers, Google Workspace Add-ons work across all platforms, including mobile devices. This cross-platform compatibility makes them a superior choice for extending functionality in Google Workspace. You can also gain users for free by listing on the Google Workplace Marketplace.
Multi-sources Retrieval-augmented generation
By retrieving information from diverse databases, knowledge graphs, or APIs, Multi-Source RAG allows models to generate more accurate, contextually enriched, and up-to-date responses. The paper outlines the architecture for combining various data sources (PDFs, Websites, APIs, etc), strategies for handling conflicting information, and techniques to optimize real-time performance.
How to build a Custom GPT
Custom GPT are ChatGPT plugins. You can make Custom GPT without coding. However, you can also code a backend API that interact with the Custom GPT. Million of ChatGPT users are browsing the GPTs Marketplace everyday. Not only it adds AI capabilities to your project, it can give your product exposures.
RAG vs Fine-tuning: which one is better for your AI project?
RAG integrates external knowledge sources to improve performance by retrieving relevant documents during inference, making it ideal for tasks that require dynamic, up-to-date information. In contrast, Fine-tuning adapts pre-trained models to specific tasks through supervised learning, optimizing performance within static, well-defined domains. We analyze the trade-offs in terms of data requirements, deployment flexibility, cost, and use cases.
Sponsored by
