Early work · 2024
LangChain over product admin guidesArchived
An early retrieval experiment: ingest a vendor's administration guides and answer configuration questions from them.
One of the first retrieval-augmented builds in the lab, from 2024: load long PDF administration guides, split and embed them, and let an engineer ask configuration questions in plain language. Crude by today's standards, but it established the pattern that later became the client knowledge base.
Related work
Athena — AI command center
Lab & internal
ClaudeOpenAIOllama CloudLocal Ollama (Jetson)+6
A private, always-on console that triages email across inboxes, monitors AI and breach news, tracks weather, and runs the company back office — with four interchangeable LLM providers behind one router.
Vectorization pipeline on Kubernetes
Early work
OpenAI embeddingsKubernetesMongoDBPython
A DevOps exercise pairing OpenAI embeddings with MongoDB on a Kubernetes cluster — the first time the lab ran a model-backed service as real infrastructure.