Advancing multimodal and agentic AI: systems, storage & scalability

Description

Developing cross-platform AI applications presents unique challenges, especially when aiming to provide consistent experiences while leveraging native capabilities. This session explores the architectural decisions behind Bodhi App, an open-source platform that enables users to run LLMs locally. I’ll share our experience using the Tauri framework—combining Rust’s performance with web technologies—to create a foundation for multi-platform support while currently focusing on macOS.

I’ll demonstrate how we structured our application to separate platform-specific code from core functionality, enabling future expansion to Windows and Linux, and possibly Android and iOS. The presentation will cover our integration with llama.cpp, explaining how we handle the communication between our Rust backend and the inference engine, while using React and modern web technologies for the user interface. This architecture allows us to deliver a lightweight application that utilizes native webviews rather than bundling entire browser engines, resulting in better performance and smaller distribution packages.

Takeaways

1. Practical insights into using Tauri for cross-platform development, including how to structure applications for future platform expansion while initially targeting a single platform.

2. Strategies for integrating AI inference engines like llama.cpp with modern application architectures that balance native performance with cross-platform compatibility.

Target Audience

This session will benefit developers working on cross-platform applications, particularly those interested in AI integration, Rust-based architectures, or alternatives to Electron. Software engineers looking to understand how to leverage web technologies without sacrificing native performance will find practical approaches they can apply to their own projects.

Bio

Amir is the founder of Bodhi App, an open-source platform for local LLM inference. With over 18 years of experience in software architecture, he specializes in developing applications that bridge high-performance backend systems with accessible user interfaces. His current work focuses on making advanced AI technologies available to everyone while preserving privacy and eliminating ongoing API costs.