Secure LLM Infrastructure

Private LLM serving stack from server to mobile:

• GPU inference — Ollama-based model serving with automated scheduling
• Authenticated API — HMAC authentication over WireGuard VPN
• Mobile client — Android app with certificate pinning and QR-code provisioning
• Benchmarking — Automated HumanEval, MMLU, GSM8K evaluation across models
• GPU comparison — Speed benchmarking across GPU generations
• Model discovery — Automated ranking and selection of new models