About
A side project that got out of hand.
By day I'm a product manager at Together AI, which means I spend a lot of time thinking about inference infrastructure, model benchmarks, and token pricing. Very grounded. Very left-brained.
I also pull tarot cards most mornings and have strong opinions about my rising sign.
For a long time those two sides of my life existed in completely separate tabs. I didn't really talk about the tarot thing at work, and I didn't think too hard about the AI thing during my more witchy moments. This was probably fine.
But I kept thinking: what would it actually look like to merge them? Not in a corporate “AI-powered wellness” kind of way. Just, what if I used the models I work with every day to build something I actually believed in?
This is that. A real tarot app, built with real AI, made by someone who uses both sincerely and is fine being judged for it.
“You may sense the truth before you fully understand it. You may feel closure before you see visible results. You may know something must change — even if you cannot yet explain why.”
— Divine Feminine Energy, The Lunar Eclipse in Virgo
Built with Together AI
Every card drawn, every reading delivered, every question asked — powered by models running on the Together AI inference platform.
Card art generation
FLUX.1-schnell
Black Forest Labs
Every tarot card you see is generated fresh for your reading. FLUX.1-schnell renders a unique 512×768 illustration for each card — styled to your chosen deck aesthetic — in under two seconds.
One image per card drawn. Each reading generates its own original art.
Personalized card faces
FLUX.1-kontext-pro
Black Forest Labs
When you upload a photo, FLUX.1-kontext-pro uses image-to-image generation to weave your likeness into the human figures of each card. Your face appears in the Empress, the Hermit, the Lovers.
Optional. Falls back to FLUX.1-schnell if unavailable.
Reading & follow-up questions
Llama 4 Maverick
Meta
The heart of the oracle. Llama 4 Maverick powers the reading engine — interpreting each card in parallel, synthesizing the full spread into a unified narrative, and generating the personalized follow-up questions before the cards are drawn.
Structured JSON output for consistent reading format. Parallel card calls for speed. Temperature 0.85 for readings, 0.8 for questions.
The Platform
All models run on Together AI's serverless inference platform — no infrastructure to manage, no GPUs to provision. One API key, unified billing, sub-second cold starts.