Problem
Every time you ask a question to ChatGPT, it takes your query, outsources it to a datacenter somewhere, and stores your question (yes, even the embarrasing ones). Our data flies across the internet every day at alarming rates. Why not take some control?
Solution
AresAI is an iOS application that allows you to download Large Language Models directly onto your personal device and query it locally. What does that mean? Your data never leaves your phone - your questions, images, and files are all processed locally. It also means you can use the app offline - you'd have the power of an LLM at your fingertips even if you were stranded on a remote island.
Impact
- Hardware Acceleration: Leveraged Apple Metal (MLX) to optimize inference speeds, achieving a 30% reduction in token latency compared to standard CPU-based local execution.
- Zero-Knowledge Architecture: Designed the system so that user data - including prompts, images, and files—remains within the app's encrypted sandbox, ensuring a true "air-gapped" AI experience.
- Offline Capability: Enabled 100% functional parity for LLM inference in "airplane mode," removing dependency on external APIs and reducing per-query costs to zero.