OpenRelay

OpenRelay is a distributed GPU inference platform. I owned the product and the brand: infrastructure dashboards, model deployment, developer tooling, onboarding, and the visual identity.

Before: an earlier OpenRelay dashboard exposing raw infrastructure across an empty overviewAfter: the operator dashboard simplified around infrastructure health, spend, and status
The OpenRelay sign-in screen: Google, GitHub, and email sign-in beside a green hero reading “Deploy globally. Scale instantly.”OpenRelay brand mark: a metallic interlocking logo centered on blackThe OpenRelay marketing site, “The inference delivery network.”, shown on a MacBookA black rooftop billboard over a brick building, reading “The only thing more expensive than SF rent is your inference bill.”
A cream billboard on a parking garage at dusk, reading “All my homies hate cold starts.”Poster: a grid of GPU racks drawn as clusters of small squares, brightening row by row, on blackPoster, “Not quite idle”: a 24-hour bar chart of GPU usage on black, captioned “On-demand GPUs that sleep when you stop paying. Billed by the hour — no commitment, no quota, no reservation.”Poster, “Let it compute.”: an isometric wireframe slab on black with a cloud of characters scattering above it and settling into a dense field beneathPoster: a crumpled thermal receipt on black — OpenRelay usage receipt, H100 and A100 GPU-hours, total 4.75, commitment none, paid in usage creditsPoster: three dot-matrix waveforms rippling across black between registration marks, footed “On-demand cloud GPUs”Poster: black squares at the ends of radiating wireframe cones, meeting at a point, footed openrelay.incPoster: concentric rings of code and dots on black around a dark center, footed openrelay.incPoster: three bands of white vertical bars over a soft grayscale gradient, footed openrelay.inc