I spent 30 years running systems where getting it wrong had real consequences — hurricane forecast models that determine evacuation orders, climate simulation clusters at national labs, the computing infrastructure that confirmed the Higgs boson at CERN. Systems where a bad NIC or a mis-configured NUMA topology costs more than money.
When LLMs became genuinely useful, I hit the same problem everyone hits: vendor lock-in, privacy risk, and the cognitive overhead of managing different SDKs and API keys for different providers. I wanted a single endpoint I could trust, that ran locally when I needed privacy, routed to cloud when I needed scale, and remembered context across everything.
Dream-Weaver started as my personal AI infrastructure. It runs in k3d on my home Alienware server, handles routing across local Ollama models and cloud APIs, stores memory in pgvector, and serves as the backbone for every AI tool I use daily. After two years of running it in production for myself, I decided to open it up.
— Kris Howard, founder
Dream-Weaver is a product of Poser8 LLC, based in Salt Lake City, Utah. Poser8 builds infrastructure software for teams running serious AI workloads — local, cloud, or hybrid.