Assortment Hub
Leading a new product improving how retail operations collaborate across supply chain, catalog, and external retailers.
I am an engineer who likes consequential systems, explicit boundaries, and tools that make difficult work feel calmer. Lately that means agentic software, model governance, and a new product boundary for retail operations. Earlier it meant Elixir platforms, Kubernetes before it was fashionable, Ruby systems, and a formative amount of time on BBSes.
Read the resume / PDFLeading a new product improving how retail operations collaborate across supply chain, catalog, and external retailers.
Composable agent framework + assistant inside SCHIP: reusable tools, citation generation and verification, progress during long tasks, reviewable automation proposals.
After six years helping lead the core SCHIP team, returned to an IC role by choice in 2026.
Early projects exploring how agents choose models, follow policies, and leave evidence of their decisions.
A governed, OpenAI-compatible model gateway that turns routing, policy, retries, and receipts into reviewable contracts instead of scattered agent logic.
A self-hosted safety layer for agents, exploring secrets isolation, destination policy, tool controls, channel routing, and auditable execution.
An evidence-oriented framework for selecting models with workload-specific Pareto frontiers across quality, cost, latency, and resource use.
Historical telemetry and trend views for Phoenix LiveDashboard.
Selected to lead a high-profile new product after six years of technical leadership on the core SCHIP supply-chain platform.
Led an enterprise product and team through scale, an acquisition, and platform migration.
Built product engineering for an internal startup; i18n and SEO.
Product work including originating a JSON-HAL hypermedia API and open-sourcing supporting tools.
Product, engineering, fundraising, partnerships.
Software R&D for Comanche and Future Combat Systems.
When an agent changes business state, a person should be able to inspect the proposal and say no — even if that adds a step.
Good explanations matter; verifiable sources matter more.
Step into management when the team needs it; return to hands-on work when that fits.
Publish early experiments without selling them as finished products.
Engineering leadership, agent systems, and products where architecture and human judgment both matter.
brian@glusman.megithub.com/bglusman