2025
Flashy OS
Two-sided SaaS — booking marketplace, DAM, and AI workspace
Two-sided platform for content production. Businesses booked photoshoots; teams managed assets, edited with AI, and shipped to clients via white-labelled share links. Used daily by the team and clients until the agency sunset.
Problem
The agency ran on inboxes and spreadsheets. Bookings came through one channel, photographer assignments through another, edits over Slack, deliverables in expiring Dropbox links. Clients waited days to see versions. We needed one system: customers book, we produce, they download — without the seams.
Role
PM'd it end to end. The product spanned a booking marketplace, an internal admin pipeline (assignment + post-production), a multi-tenant DAM with semantic search, AI image and image-to-video editors, version + comment threads, a Meta social publisher, role-based access, and white-label share pages. As PM I wrote the PRDs and user stories, prototyped flows in v0, set up Amplitude to track product analytics, and prioritized the roadmap with a cross-functional team of 7 across engineering, design, and GTM. I onboarded the team and clients, then supported it daily.
Outcome
Used daily for ~14 months until the agency sunset. Semantic search and the AI editor cleared a long-running bottleneck — clients stopped asking for 'the version with the blue mats' because they could just type it. Lessons: I over-engineered the invoicing module before talking to the people who'd use it, and under-engineered the deliverables module that turned out to be where the team actually lived. Next time I'd ship the rough version of the right thing instead of the polished version of the wrong thing.
Stack
- Vue 2
- Quasar
- Hapi
- Node.js
- PostgreSQL
- pgvector
- Pinecone
- OpenAI
- Replicate
- AWS Rekognition
- Stripe
- Auth0
- Meta Graph API
- S3
- SQS
- AWS Lambda
- Terraform
Next project
Flashy Models →
Vibecoded model selection tool