Sage Forge

Inspiration Wearables today excel at collecting health data but rarely help users improve their daily lives. Most devices stop at dashboards, leaving users to interpret metrics and decide what to do next. We wanted to build something fundamentally different: an AI companion that understands context, learns continuously, and proactively helps users make better decisions. Our vision was to transform a smart ring from a passive sensor into an intelligent software platform that people can interact with every day.

What it does Lumi is an AI-native wearable platform that combines multimodal sensing with autonomous AI agents. The platform continuously fuses physiological signals, motion, environmental context, conversations (with permission), user interactions, and historical behavior to build a long-term understanding of each user. Instead of presenting raw metrics, Lumi reasons over this information and delivers personalized coaching, emotional support, productivity assistance, relationship insights, and contextual recommendations. The platform is extensible through an AI Skill ecosystem, allowing developers to create new capabilities while users personalize Lumi with skills that match their lifestyle.

How we built it Lumi consists of four major software layers: Multimodal Intelligence Layer – Collects and processes wearable sensor data, phone context, voice, images, and user interactions. Long-Term Memory & Personalization – Maintains evolving user profiles, preferences, habits, and behavioral history for personalized reasoning. Agentic AI Platform – Multiple specialized AI agents collaborate to understand context, plan actions, retrieve knowledge, and proactively assist users. Developer Ecosystem – An extensible framework that enables third-party AI skills and future integrations. The wearable serves primarily as an always-on sensing device, while the majority of the intelligence resides in the software platform.

Challenges we ran into Building Lumi required solving problems beyond traditional wearable applications. Combining heterogeneous sensor streams into a unified user context. Designing long-term memory that improves personalization without becoming repetitive or intrusive. Balancing proactive assistance with respecting user privacy and control. Managing latency and battery constraints while supporting continuous AI experiences. Creating a software architecture flexible enough to support future AI skills and developers. Accomplishments that we're proud of Built an AI-first wearable platform instead of another health dashboard. Developed a multimodal reasoning engine capable of understanding rich personal context. Designed an extensible AI Skill architecture for future third-party development. Successfully integrated continuous sensing with proactive AI experiences. Established a business model centered on recurring software value rather than one-time hardware sales.

What we learned One of our biggest insights was that hardware alone does not create long-term engagement. Users return because software continuously becomes more useful as it learns about them. We also learned that personalization requires more than LLMs. Building a truly helpful AI companion depends on long-term memory, contextual reasoning, and carefully orchestrated agents working together—not just generating responses. Finally, we realized that the wearable should disappear into the background. Users care about meaningful outcomes, not raw sensor data.

What's next for Lumi Our roadmap is focused on expanding Lumi into a complete AI ecosystem. Launch additional AI agents for productivity, wellness, relationships, and personal growth. Open the AI Skill SDK for developers to build and monetize new capabilities. Expand multimodal reasoning with richer environmental and contextual understanding. Improve long-term personalization through continual learning while maintaining strong privacy controls. Grow Lumi into an AI operating system for everyday life, where the wearable becomes the always-on interface to an evolving ecosystem of intelligent services.

Built With

• airi
• elevenlabs
• gpt
• hume
• stt
• three.js
• tts