Building local-first AI runtimes, Wasm-sandboxed MCP tooling, and distributed media systems.

Website · LinkedIn · Email · GitHub

• Languages & Systems: Rust, TypeScript, Python, C++ (Embedded/ESP32)
• AI & Agentic Tooling: Model Context Protocol (MCP 2024-11-05), GGUF/safetensors, Unsloth, VRAM Budgeting, Wasmtime / Extism (WASM)
• Web, API & Backend: Nuxt 4, Vue 3, Bun, Nitro v3, Axum, Docker Swarm, Redis, CockroachDB
• Media & Streaming: FFmpeg, WebCodecs, LiveKit SFU, HLS, Cloudflare R2 / AWS S3

1. Rune Ecosystem (AI Runtimes & MCP Tooling in Rust)

• Rune Infer: High-performance, single-binary AI model gateway unifying local GGUF engines and cloud APIs with pre-flight VRAM budgeting, LRU model unloading, and zero-crash isolation.
• Rune Kit & Tools: Rust runtime host and package manager for the Model Context Protocol (MCP), featuring secure, memory-isolated wasm32-wasip1 Extism sandboxes and Stdio JSON-RPC multiplexing.
• UnAi: Unified TypeScript adapter standardizing text, streaming, structured JSON schemas, and tool execution across 7 providers and 56+ models.

2. Enterprise Backend & Media Infrastructure (Modest Human Brands)

• MMedia: Event-driven, containerized media microservice orchestrating background job queues, hardware-accelerated DASH transcoding via FFmpeg, and LiveKit WebRTC-to-HLS live broadcasting.
• MStudio: High-performance desktop workstation (Rust, Tauri v2, Vue 3) featuring P2P asset synchronization via Iroh QUIC streams, multithreaded FFmpeg bulk watermarking, and an Axum PNA cryptographic bridge.
• MDoc & MConnect: Cryptographic PDF signing microservice (Nitro v3) and omnichannel engagement gateway orchestrating WebRTC/SIP telephony, WhatsApp, and anti-bot email telemetry.