Preston Bernstein

I’m Preston Bernstein, a full-stack developer based in Atlanta with over ten years of experience building and optimizing web applications, and the infrastructure they run on.

Most of what I write about comes out of my own home lab: a Synology NAS, a desktop with one much-fought-over GPU, a retired laptop running Proxmox VE, and a UniFi network with Pi-hole DNS filtering underneath all of it. On top of that hardware I run self-hosted media automation, photo backup, knowledge-graph pipelines built on LightRAG, local LLM inference through Ollama, and a growing amount of AI agent tooling — Claude Code pipelines that spec, build, review, and deploy software with me supervising rather than typing.

The blog is where the debugging sessions and design decisions get written down honestly, including the parts that failed or stayed unresolved. Posts cover home-lab infrastructure and networking, self-hosted service placement and observability, and practical AI/LLM engineering: agent workflows, local inference, retrieval pipelines, and cost control.

A few places to start:

• What a $364 Claude Code session taught me about running agents unattended
• Three failure modes wearing one name: running concurrent Claude Code agents
• Tuning LightRAG ingestion concurrency against a rate-limited Gemini API
• Not every Docker container belongs on the NAS

You can read more about me, browse the blog, see my projects on GitHub, or connect on LinkedIn.