Carlos Q.

Some businesses can't use AI the way most freelancers deliver it. If you're in healthcare, legal, or financial services, your data can't quietly flow through third-party APIs — and a ChatGPT wrapper won't survive your security review. I design and deploy local-first AI infrastructure: LLM pipelines that run on hardware you own, or a hybrid you control — built as production Python with automated tests, not notebooks. My systems are designed to support HIPAA-regulated workflows; privacy guarantees apply to deployed systems (prototyping may use synthetic data). I run frontier-class open-weight models on dedicated hardware (RTX Pro 6000 Blackwell, 96GB VRAM) and can show you exactly what runs locally, at what speed, and at what quality. Where I help: • Document intelligence — purchase orders, invoices, contracts, and scanned records turned into structured data pipelines with human review gates • Private RAG — chat-with-your-documents over your own knowledge base, on-prem or hybrid • Agentic workflow automation that replaces manual ops (intake, reporting, back office) with audit trails and human approval steps • Open-weight model work — quantization, evaluation, and serving across inference engines including vLLM, llama.cpp, Ollama, and ninfer How I work: scoping session → fixed-scope prototype → hardening, deployment, and documentation, so your team owns the system after handoff. I also work fluently in modern AI-assisted development workflows (Claude Code, Codex, Pi, opencode), including fully local agent setups. Native English and Spanish. Based in Orlando, FL, working US hours. If you need AI your IT director can approve and your compliance officer can sign off on, reach out and we'll scope it.