Skip to main content
All work

One agent platform for four industries

I built a white-label AI agent platform and deployed it for four industries: therapy, home health, pharmacists and financial advisors. Each has its own domain and branding. It took 8 days, on top of the Suna agent framework, opens in a new tab.

Industries
Therapy, home health, pharmacists, financial advisors
Length
8 days
Outcome
Went live; the client tested the market, then pivoted
The financial advisors version: its landing page, then the agent working through a research task, sped up 2 to 4 times while it works. It plans its tasks at 0:20, researches the web from 0:24, and has written its report by 0:44.

What it is

One platform, four products. Each industry has its own domain, branding, landing page and agents, and all of them run from the same platform.

  • Four industries. Therapy, home health, pharmacists and financial advisors.
  • Their own domains. Each industry runs on an independent domain, with its own branding.
  • A template for new agents. A framework for creating and deploying the agents for an industry.
  • Chat that streams. Answers stream in, and the conversation’s context is kept across turns.

What I did

The agents run on the Suna framework. My part was turning it into four deployable products.

  • Deployment as code. The AWS infrastructure is defined with CDK and deployed through GitHub Actions, with monitoring.
  • White-label customization. Branding and interface changes for each industry, across every customer touchpoint.
  • Landing pages and onboarding scripts. A marketing page for each industry joined to the main app, and scripts that configure and deploy a new one quickly.
  • Trial access. Guests can try it within usage limits, and conversions are tracked.
  • Industry-specific behavior. Each industry’s agents behave differently, with their own knowledge bases and workflows.

How it fit in 8 days

It came together in 8 days through pre-built framework components and automated deployment pipelines.

What happened next

All four versions went live. The client marketed them with Meta ads and collected a lot of metrics and data. Based on what they learned, they decided not to pursue any of the four and pivoted to something different.

Tech stack

Frontend

  • Next.js
  • Tailwind

Backend

  • Python and FastAPI
  • Supabase

AI processing

  • Suna agent framework
  • Models from OpenAI, Anthropic, OpenRouter and others

Infrastructure

  • AWS CDK
  • GitHub Actions

Tell me what you’re trying to ship.

Free · no prep needed · proposal in 48h