Bloomkin

Home Page

Inspiration

We've all been stuck on a skill with no one to turn to. Existing platforms let you game the system by searching for friends. We wanted the best mentor to find you based on what you want to learn. Bloomkin — like a flower blooming — is about growing with the right guidance. But a garden thrives together: by sharing your knowledge, you earn credits to learn even more. To help another bloom is to water your own roots.

What it does

AI-powered mentor matching. Describe what you want to learn, get matched with the best mentors by skill fit and reputation. Book a session, get a Google Meet link automatically, validate it happened, leave a rating. Mentors earn credits from cumulative stars — every 5 stars = 1 credit. Pure reputation economy.

How we built it

Next.js 15 — App Router, Server Actions
Supabase — Auth (Google OAuth), database, RLS
Gemini 2.5 Flash — AI mentor matching with keyword fallback
Google Calendar API — auto-generates Meet links at booking
Claude — background fraud detection on ratings

Challenges we ran into

Google Calendar OAuth server-side setup with conference data (Meet links) required specific API flags that weren't obvious
Preventing abuse without breaking normal flows: one rating per session, rating-before-rebook enforcement, partial unique indexes on bookings
Gemini's 20 req/day free tier forced us to build a keyword fallback so search never breaks during demos

Accomplishments that we're proud of

A fully working end-to-end flow: search → AI match → book → Google Meet invite → validate → rate → credit earned → notification. Built and shipped in a hackathon weekend with real third-party integrations throughout.

What we learned

Designing for abuse prevention from day one changes everything — RLS policies, partial unique constraints, and session validation logic need to be planned before you write a single feature. AI works best when it runs silently in the background.