Qytsol product

Exprty

A creator platform where coaches, teachers, consultants and practitioners share one link and sell paid 1:1 calls, group sessions, cohort courses and digital products to their audience. Bookings, payments, video rooms, downloads and payouts all run underneath, and payments split at capture, so the platform never holds the creator's money.

Role
Product design, architecture, full build and ongoing operation.
Timeline
January 2026 to present
Platforms
Web, Responsive mobile
Status
Live in production

Visit exprty.com (opens in a new tab)

The Exprty home page, headed "One link to earn from your expertise", beside a sample creator page selling a 1:1 call, a group session, a four-week cohort and a PDF guide

The problem

Plenty of people have knowledge their audience would pay for. A YouTuber’s viewers want a live workshop. A LinkedIn mentor gets the same “can I ask you something?” message every week. A teacher has notes and recordings that could be a course. A consultant could sell an hour of judgement to people who follow their work.

Turning that into income means stitching together a scheduler, a payment link, a video tool, somewhere to host files, and a way to issue tax invoices and get paid out. Each piece is a separate account and a separate bill, and none of them know about each other. A cancelled group session becomes a dozen manual refunds. A course becomes a spreadsheet of who paid for which week.

Exprty collapses all of that into one link. A creator applies, sets up their offers and shares a page. Their audience picks a time or a product and pays. Bookings, video rooms, downloads, invoices and payouts happen without the creator running any of it.

What we built

Exprty is a two-sided platform where creators sell four kinds of offering from a single public page, all settled by one booking, payment and scheduling engine.

Four formats, one engine

  • 1:1 calls: one person books one slot, from 15 to 60 minutes, at the creator’s rate. The creator accepts each request, or turns on auto-approval so a paid booking confirms instantly. Instant Help offers on-demand calls with whoever is available now.
  • Group sessions: a date, a price per seat and a seat limit. Several people each pay for a seat and join one video room.
  • Cohort courses: a series of live group sessions sold as one seat. The buyer pays once and attends every session.
  • Digital products: guides, templates, e-books, recordings or resource links, sold at a fixed price or given away free.

For creators

  • A readable public link to share on Instagram, LinkedIn, YouTube or WhatsApp, with rich previews when it is posted, a share kit and a “book me” badge
  • An application and review step, so a page goes live only once it is approved
  • Per-offer pricing, recurring weekly availability in their own timezone, and the option to use their own Zoom or Meet link instead of the default room
  • Bookings, seats, orders, earnings and analytics in one dashboard
  • Bank onboarding to a payment linked account, which gates whether they can sell at all
  • Automatic invoices, generated as PDFs

For buyers

  • Pick a time first and sign up only at the last step
  • Book a 1:1 call, take a seat in a group session or cohort, or buy a product
  • Pay per session in rupees, with no subscription
  • A structured session brief built from free text, sent ahead of the call
  • Live video in the browser, in-app messaging, and post-session recaps
  • Purchases, session history, invoices and feedback in one place
  • A first-session guarantee: if a first session is not useful, it is refunded in full within 24 hours, once per account

Running the platform

  • Admin tooling for application review, approvals, bans, refunds and a money view backed by daily payment reconciliation
  • Transactional email, web push notifications, calendar invites and one-time passcodes
  • Scheduled jobs for reminders, post-session nudges, attendance, expiring unpaid bookings and seats, and sweeping orphaned payments, with every run logged
  • Error monitoring and structured retry around every database call

Engineering challenges

Moving money without ever holding it

Exprty splits every payment at the moment of capture. Each creator onboards to their own linked account with their bank details, and every booking attaches a transfer to that account with a hold until shortly after the session. The creator’s share goes directly to them and releases automatically. The platform never takes custody of it at any point.

That design is deliberate, and the reason is regulatory as much as technical. The moment a marketplace holds the seller’s share between the sale and the payout, it is a custodian of other people’s money, which is precisely what payment aggregator rules exist to govern. Splitting at capture removes the question entirely rather than answering it.

Building it that way puts the difficulty in three places:

  • Sellability is gated on payout readiness. A creator who has not finished bank onboarding cannot be booked or bought from at all, and the rule is enforced in the queries that build the public listing as well as at checkout, because a sale that cannot be paid out is worse than no sale.
  • Refunds run in reverse. The transfer to the creator is unwound before the buyer is refunded, never after, so the two can never both be outstanding. The first-session guarantee runs through this same path, which is why it could be offered at all: an unconditional refund promise is only safe when the unwind is automatic and ordered.
  • Reconciliation is automatic. Scheduled jobs expire unpaid bookings, resume interrupted payments and sweep orphaned ones, and a daily job checks the payment provider against the database, so a dropped webhook or an abandoned checkout never leaves a booking and a payment disagreeing about what happened.

Four formats, one booking engine

1:1 calls, group sessions, cohorts and products look like four products to a creator and are one system underneath. A group seat is an ordinary booking with its own slot, so payment, the split, refunds and invoices all reuse the 1:1 pipeline unchanged. A cohort seat is one booking whose slot runs from the first session to the last, which means the creator’s payout hold naturally releases only after the final session has happened. Products reuse the same order, split and refund machinery without a slot at all.

The alternative, a separate flow per format, would have multiplied the payment code by four, and that is exactly the code that must never diverge. Keeping one engine means a fix to the refund order or the payout hold lands everywhere at once.

Selling seats without overselling them

Group sessions and cohorts have a seat limit, and the last seat is exactly when two people are most likely to check out at the same moment. Each seat purchase locks the session row before counting, so two buyers can never both see one seat left and both take it. Unpaid seats count toward capacity while checkout is in progress and expire on a timer, so an abandoned payment cannot hold a seat forever.

When a creator cancels a whole session or cohort, every paid seat is refunded server-side in one idempotent job that is safe to re-run halfway through, rather than by the browser looping over refunds and hoping the tab stays open. Meeting links live apart from the session itself, so someone holding an unpaid or cancelled seat can never read the room link.

Delivering files only to people who paid for them

Digital products live in a private storage bucket. A buyer never sees a permanent URL: each time they open a purchase, the server checks the order and issues a signed link that expires in 60 seconds. The first open is recorded before the link is handed over, and that moment closes the self-serve refund window, since a file that has been downloaded cannot be returned. Refunds after that point go to admin review instead of being granted automatically.

Getting timezones right, not approximately right

Availability is published by creators in their own timezone and consumed by buyers in another. Every time is stored and compared in UTC and displayed in local time, which is the standard advice and not the hard part.

The hard part is that the single most common way to derive a calendar date in JavaScript is silently wrong. Taking an ISO string and slicing the first ten characters converts to UTC first, so for any timezone ahead of UTC it returns the previous day for a large part of every evening. The same trap sits in parsing: a bare date string is read as UTC, not local.

Both patterns are banned in this codebase and replaced by a small set of shared helpers that build and parse dates from local components. It is an unglamorous constraint, and it is the difference between a booking calendar that works and one that drops a day for half its users after 5:30pm.

An AI layer that is allowed to fail

Exprty uses AI to turn a buyer’s free-text notes into a structured session brief, to write post-session recaps, and to explain why a given expert fits a search. Matching runs on vector similarity over embedded expert profiles; the generative features run on a hosted model from an edge function.

The design rule is that no AI feature is ever on the critical path. Every one of them has a defined non-AI fallback: a brief falls back to the buyer’s plain notes, matching falls back to keyword search, and explanations render as nothing at all rather than blocking the results. Recaps are generated once and persisted on the booking, so repeat views cost nothing and produce the same answer. Someone who arrives while the model is rate-limited still books their session.

Architecture

Two paths out of the browser, chosen deliberately. Row-level security lets the client read and write its own data directly, and everything that holds a secret or moves money runs server-side instead.

Exprty system architecture A React single-page application reads and writes PostgreSQL directly under row-level security, and calls Deno edge functions for anything requiring a secret. The edge functions talk to PostgreSQL and to five external services: Razorpay Route for split payments, web push for notifications, Jitsi, Zoom and Google Meet for meetings, Resend for transactional email, and Google Gemini for the generative AI features. Web client React SPA, creator and buyer Edge functions Deno. Anything needing a secret PostgreSQL RLS, pgvector, file storage Auth and realtime Sessions, live updates Razorpay Route Web push Jitsi, Zoom, Meet Resend email Google Gemini privileged calls direct reads, RLS
The browser reads and writes PostgreSQL directly, with row-level security rather than a hand-written API layer deciding what each role can see. Anything holding a secret, moving money or calling a third party runs in an edge function instead.

Technology

Frontend

  • React 19
  • TypeScript
  • Vite
  • Tailwind CSS
  • shadcn/ui
  • React Router

Backend

  • Supabase
  • PostgreSQL
  • Deno Edge Functions
  • Row-Level Security
  • pg_cron

Data

  • PostgreSQL
  • pgvector
  • Supabase Storage
  • Realtime subscriptions

Payments

  • Razorpay
  • Razorpay Route

Communications

  • Jitsi (8x8)
  • Zoom and Google Meet links
  • Resend
  • Web Push

AI

  • Google Gemini
  • Supabase embeddings
  • Vector similarity search

Platform

  • Vercel
  • Vercel Edge Middleware
  • Sentry
  • Vitest
  • fast-check

Outcome

Live, and running itself.

Exprty is live and running in production. We have designed, built and operated it since January 2026, and took it from a marketplace for engineering help to a platform any expert can sell from without rebuilding the core. The figures below are a fair proxy for the surface area involved: four selling formats, payments, payouts, scheduling, seat capacity, file delivery, invoicing, reconciliation, admin tooling and an AI layer, all of it under test.

What that surface area buys is a payment system nobody has to think about. A buyer pays once. The creator's share splits directly to their own account at the moment of capture, is held until just after the session (or the last session of a cohort), and settles automatically. Invoices generate themselves, refunds unwind the whole chain in the correct order, and a daily reconciliation keeps bookings and payments in agreement without anyone intervening. Money movement is the most expensive thing in a marketplace to get wrong, and on Exprty it runs untouched.

62 Serverless functions in production
303 Versioned database migrations
161 Test files, including property-based tests
4 Ways to earn, sold from one link

The product

We can build your product the same way.

From first conversation to a product in front of real users.