I ship production systems across Python, Rust, TypeScript, and Go. 3 published research papers. 9 models on Hugging Face. From fine-tuning LLMs to building air-gapped wallets in Rust.
Deep specialization at the intersection of AI, distributed systems, and blockchain infrastructure.
End-to-end ML pipelines from data curation through fine-tuning to quantized local deployment. Multi-provider fallback architectures for production reliability.
Decentralized GPU inference benchmarking, video processing pipelines, and edge compute architectures. Performance-critical Rust services.
Air-gapped cold wallets, DCA trading bots, cross-chain DeFi research, and on-chain attestation systems. Secure key management in Rust.
CLI toolkit suites, MCP servers for AI orchestration, CI/CD pipelines, and deployment automation across Cloudflare and GitHub Actions.
DeepSeek-R1 fine-tune for distributed systems and DeFi. 4 GGUF quantizations with Ollama Modelfile. Trained on Apple Silicon with MLX LoRA.
Distributed LLM inference on decentralized GPU networks. Benchmarked against centralized cloud at scale.
GPU-accelerated video processing pipeline on distributed compute. NVENC/CUDA encoding with broadcast-grade quality.
Air-gapped cold wallet for Solana. Rust secure signer with mlock/zeroize memory protection. Bootable Alpine Linux USB.
7-crate Rust video player. HLS/DASH streaming, GStreamer, WASM, Tauri desktop, CLI, audio fingerprinting, Python bindings.
DCA bot for Solana using Jupiter v1 swap API. Live mainnet transactions verified. Time-weighted average price execution.
Distributed LLM Inference on Decentralized GPU Networks
GPU-Accelerated Video Processing on Distributed Compute
Air-Gapped Cold Wallet Architecture for Solana
AI/ML engineer and distributed systems builder. Founder of Purple Squirrel Media.
I've spent 5+ years at the intersection of AI infrastructure, distributed systems, and blockchain. I build things that work in production — from LLM fine-tuning pipelines to air-gapped hardware wallets.
My research on decentralized GPU inference showed a 47% cost reduction versus centralized cloud. I train models locally on Apple Silicon, publish them on Hugging Face, and ship quantized endpoints for local deployment. Everything is MIT licensed and open source.
I write Python, Rust, TypeScript, Go, and Ruby. I've built MCP servers, CLI toolkits, DCA trading bots, and 7-crate Rust video players. I work best on hard infrastructure problems.
Published 9 models, 3 research papers, and 3 datasets covering distributed inference, GPU video processing, multichain DeFi, and air-gapped Solana wallets. All MIT licensed with GGUF quantizations and Ollama Modelfiles.
View Collection →The next wave of developer productivity isn't about writing code faster. It's about AI that understands your entire codebase and can act on it. MCP is the protocol that makes this real.
Read more →Model Context Protocol is quietly becoming the standard for AI-to-software integration. I've built 5+ servers on it and here's what I've learned about production MCP development.
Read more →Why I built a USB-based cold wallet that keeps keys in RAM only during signing — and what it taught me about security UX and Rust memory protection.
Read more →Looking for AI/ML infrastructure, distributed systems, or blockchain engineering roles. Open to full-time remote positions.
Schedule directly on my calendar