# Building Local AI with Homebrews Full-Stack Approach > [! note]- > The content of this page is generated by audio/video transcription and text transformation from the content and links of this source. Source: [https://fosdem.org/2025/schedule/event/fosdem-2025-5059-building-local-ai-with-a-full-stack-approach/](https://fosdem.org/2025/schedule/event/fosdem-2025-5059-building-local-ai-with-a-full-stack-approach/) <video src="https://video.fosdem.org/2025/k1105/fosdem-2025-5059-building-local-ai-with-a-full-stack-approach.av1.webm" controls></video> ## Summary & Highlights: **Introduction to Local AI with a Full-Stack Approach** In this session, Rex Ha from Homebrew discusses the importance of a full-stack approach in building local AI systems. The talk highlights hardware compatibility, multimodal model optimization, and efficient inference on edge devices. **Deploying AI to Over 1 Million Devices** Rex shares insights from deploying local AI to over a million devices, focusing on optimization hurdles, architectural trade-offs, and hardware requirements. The session emphasizes the impact on both developers and users. **User Feedback and Continuous Improvement** The talk explores how integrating user feedback loops and collecting RLHF data can continuously improve model performance, offering more effective solutions. **Technical Insights and Challenges** Rex discusses the technical aspects of running AI models locally, including the challenges of training and deploying large language models on consumer hardware, and the importance of optimizing for smaller, smarter models. ## Importance for an eco-social transformation The session is significant for eco-social transformation as it addresses the efficient deployment of AI on local devices, reducing energy consumption and reliance on large data centers. This approach can empower communities to have more control over their data and AI applications, enhancing privacy and customization. Eco-social designers can leverage tools like JAN and Cortex for community-focused AI solutions. Challenges include ensuring equitable access to necessary hardware and overcoming technical barriers in optimizing models for local use. ## Links [Video recording (AV1/WebM) - 138.3 MB](https://video.fosdem.org/2025/k1105/fosdem-2025-5059-building-local-ai-with-a-full-stack-approach.av1.webm) [Video recording (MP4) - 637.5 MB](https://video.fosdem.org/2025/k1105/fosdem-2025-5059-building-local-ai-with-a-full-stack-approach.av1.mp4) [Video recording subtitle file (VTT)]() [Chat room(web)]() [Chat room(app)]()