# Milo

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Arjun S](https://cerebralvalley.ai/u/GS3)
- **GitHub:** https://github.com/83Gh0st/milo
- **Demo video:** https://youtu.be/96NygsvKZII
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1806

Milo is a personal AI companion I built to help children with dyslexia, people with speech and language disabilities, and individuals who experience difficulties with everyday communication. For a child with dyslexia, reading a paragraph, understanding written instructions, or expressing an answer in writing can be difficult even when they understand the underlying idea. For someone with a speech or language disability, communicating a simple need or explaining what they are thinking may not always be possible through speech. Some people communicate more comfortably using pictures, symbols, gestures, or augmentative and alternative communication (AAC) tools. However, most digital assistants still expect users to type, speak, and interact through conventional interfaces. I wanted to explore what an AI companion would look like if it could accommodate these different ways of communicating.

Milo brings together reading and comprehension support, visual communication, voice interaction, image understanding, and everyday assistance in one mobile application. A user can ask Milo to simplify a difficult passage, explain an instruction in easier language, or help understand information through visual examples. Someone who has difficulty expressing themselves verbally can use pictures, symbols, and simple selections through a visual communication board to communicate their needs. Milo also supports voice-based interaction, image understanding, illustration generation, stories and assistance with everyday activities. A persistent animated character acts as the interface across these experiences, giving users a familiar point of interaction rather than presenting them with a conventional chatbot.

I developed Milo using Flutter, with a FastAPI backend deployed on Modal. Meta's AI models power its reasoning and multimodal capabilities, including speech transcription, image generation, visual understanding, and object segmentation. I wanted the application to be useful even when an internet connection is unavailable, so essential communication features and the visual communication board work locally, independently of the online AI services. The architecture also incorporates constrained AI interactions, application-side validation, user-controlled memory, and privacy-conscious handling of user data.

One of the ideas behind Milo is that users should not have to adapt themselves to technology. A child should not need to navigate complicated settings to communicate, and someone who cannot comfortably use speech or text should still have meaningful ways to interact. Milo is my attempt to bring these different forms of support together into a single companion that can make reading, learning, communication, and everyday activities easier to navigate. Built for the Meta Global AI Developer Hackathon, the project explores how multimodal AI can become more than a tool for answering questions and instead help people express themselves and participate more independently in everyday life.

## More from Meta Global AI Developer Hackathon

- [Stall to Store](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1803)
- [PerceptaHeal](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1804)
- [Ikigai Loop](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1805)
- [Mirai Climbing](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1807)
- [Ai study assistant](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1808)
- [Buildlers](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1809)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1806. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt