Case study · My own product

Talent Flow: WhatsApp intake for talent agencies

Talent Flow is my own product, built and run by me as Wolfsteno: a multi-tenant WhatsApp intake platform for talent agencies with automated conversations, structured candidate records, and a shared inbox for human follow-up.

Visit the product pageGet in touch

My role

Everything: product, design, code, infrastructure and operations. Talent Flow is my own product, built and run by me as Wolfsteno.

Tech stack

  • Angular
  • Express
  • PostgreSQL
  • Prisma
  • WhatsApp Cloud API
  • TypeScript
  • Docker

The problem and the build

Talent agencies often receive their first contact through WhatsApp, then spend time copying answers into spreadsheets and piecing together the conversation later. Talent Flow turns that chat into a guided intake: it welcomes the contact in their language, asks the right questions, stores the answers as a candidate or client request, and hands the conversation to the team when a person should take over.

Each agency gets an isolated operations portal with its own users, configurable conversation flows, candidates, requests, exports and shared inbox. A separate super-admin portal manages the platform and its tenants, while the bot backend handles Meta webhooks, message delivery and the conversation state.

The system is built as five independently deployable applications: Angular portals for agency staff and platform administration, Express APIs in TypeScript, PostgreSQL through Prisma, and a dedicated WhatsApp Cloud API webhook service. It runs on a Debian VPS behind nginx with path-filtered CI/CD deployments.

Highlights

  • A visual flow designer for building guided conversations on a canvas and testing them in a live preview.
  • A shared real-time inbox with outbound replies, delivery updates, agent attribution and detachable conversation windows.
  • Structured candidate and client-request records with search, editing, filtering and spreadsheet export.
  • True multi-tenancy: isolated agency portals, users, data and bot configurations, managed from a separate platform portal.
  • Four bot languages: English, Spanish, Ukrainian and Russian, selected as part of the WhatsApp conversation.
  • Self-hosted end to end, with PostgreSQL, nginx, systemd services and automated deployments on a Debian server I manage.

Screenshots

  • The operations dashboard, showing candidate, conversation, message and handoff totals at a glance.
  • The visual bot flow designer, with the conversation graph on the canvas and a live test preview beside it.
  • The shared team inbox, with the conversation list, the WhatsApp history, voice notes with transcripts and the contact details.
  • A candidate record, with the profile photo, contact details, messaging consent and the CV facts.
  • The CV editor, drafting a client-ready CV with AI and previewing the printed page beside it.
  • A client request with suggested candidates, each scored and explained, ready to send with their CV.
  • The AI flow designer, turning a plain-language description into a draft conversation on the canvas.