CONIFER • INFERENCE GATEWAY • 2026
Run AI locally. Route the rest. Pay less.

- Role
- Product Designer
- Timeline
- 2025 – 2026
- Team
- Conifer (YC S26)
- Skills
- Developer tools and AI product UX, Marketing site and narrative design, Systems thinking for technical products
Proof
Live prototype
Problem
Conifer routes inference to the cheapest model that still clears the work — local hardware first, cloud when needed. The product is real and shipping; the site had to explain that routing story without sounding like every other AI API landing page.
The design job was credibility for a technical buyer: make local-first routing legible, show the install path immediately, and keep the YC-stage product feeling crafted — not templated.
Process
I mirrored the live product, mapped the narrative gaps, and rebuilt the landing as a static redesign — hero, routing story, model exploration, and footer — before handing off to production.
The ASCII bonsai and marble ground became the visual anchor: a craft signal that reads developer-native without abandoning clarity for non-designer buyers.
Outcome
The redesign gives Conifer a single coherent story: install locally, route intelligently, pay less. Model exploration and pricing support the hero instead of competing with it.
Shipped as a reference redesign on the path to production — typography, spacing, and section rhythm tuned for conifer.build's technical audience.
My role
Owned
- Landing page redesign for conifer.build
- Hero narrative, routing story, and install flow UX
- Visual language — monospace craft, ASCII bonsai, marble surfaces
- Responsive layout and dark/light theme behavior
Did not own
- Gateway routing engine and model catalog backend
- CLI and console implementation
- Billing, auth, and infrastructure
Collaborators: Michael Jeffords, Conifer engineering
Key decisions
Decision 1
Feature dump vs. one routing story
Chose: Lead with a single promise — local-first routing to the Pareto frontier — then unpack install, models, and pricing in that order.
Rejected: A docs-style homepage listing every endpoint first. Engineers bounced before understanding why Conifer exists.
Decision 2
Generic AI landing vs. Conifer craft
Chose: ASCII bonsai, marble ground, and monospace type as brand signals — technical without reading like every other inference startup.
Rejected: Gradient SaaS template with stock illustrations. Fast to ship, zero differentiation in a crowded category.
Decision 3
Explain routing abstractly vs. visually
Chose: A dedicated routing diagram and model explorer sections so users see the decision path, not just the tagline.
Rejected: Copy-only explanation of cost savings. Hard to trust a routing claim you can't picture.