I specialize in building practical AI systems, agentic workflows, and adapting open models. My focus covers both ends of the LLM pipeline: fine-tuning open-source models for specialized tasks, and building the infrastructure where they connect with real environments (tool integration via MCP, multi-agent coordination, and resilient provider routing).
Coming from a hands-on background in hardware diagnostics and system administration, I approach AI with a practical mindset: prioritize reliability, low latency, and predictable execution over hype.
- 🎯 Model Fine-Tuning: Adapting open-source LLMs (LoRA/QLoRA, dataset preparation, format alignment) for domain-specific tasks and structured outputs.
- 🔌 Tool Integration & MCP: Connecting AI models to external tools, APIs, and developer environments via Model Context Protocol.
- 🤖 Agentic Workflows: Designing multi-step agent setups with tool calling, evaluation loops, and fallback logic.
- ⚡ LLM Routing & Gateways: Load-balancing across multiple providers, handling failovers, and managing quotas.
- 🛠️ Systems Foundation: Experience with both physical hardware and the operating systems running the software.
🧠 AI & LLM Systems │ Model Fine-Tuning (LoRA/QLoRA), Agentic Workflows,
│ Consensus Routing, MCP Servers, Prompt Architecture
⚙️ Languages & Runtimes │ TypeScript, Python, Node.js, JavaScript, Ruby, Bash
🗄️ Databases & Storage │ SQLite, PostgreSQL, Redis
🚢 DevOps & Infrastructure │ Docker, Linux CLI, PM2, Reverse Proxies, Cloudflare, Bare-Metal
Kitesurfing · Surf · Chess · Hardware tinkering · Testing emerging tech
Open to technical discussions, fine-tuning workflows, and AI systems architecture.