Building Dochelp.ai - From MVP to SASS

An image that representes the journey from an MVP concept app to a fully functional software as a service

hey there! thought i'd share a bit about what it's like wearing all the hats as the founder and cto of dochelp.ai.

dochelp.ai

imagine a chatgpt but for doctors. we basically connect gpt models with some medical pdfs so the AI can answer correctly and more precisely.

current frontend stack:

  • next.js
  • tailwindcss
  • shadcn/ui

current backend stack:

  • node.js
  • fastify
  • drizzle orm
  • postgresql
  • vitest
  • openai
  • langchain
  • upstash vector store

infra:

  • vercel for frontend deploy
  • railway for backend deploy
  • github with ci/cd actions
  • resend for sending messages

where we're at?

  • 2023/jan (first idea): anamnese ai generator
    • this idea came as a freelance project for me, i delivered but the client didn't go forward with the app
  • 2024/jan (pivot idea): chatgpt for doctors mvp 1
    • i showed the custom gpt assitant that you can create and upload some pdf's for this old client
    • we talked about creating an mvp and trying to get people to use it
    • created an instagram profile and got about 300 subscribers into a online form
    • put them into a whatsapp group
    • developed the mvp within 2 days, this app version consisted of:
      • one single page next.js app
      • one text input only fixed at the bottom
      • user send message would connect to langchain
      • langchain would consult upstash vector store database
      • using langchain rag to keep context consistent
      • if user refresh pages it would reset the chat messages
      • user could not access previous chats
      • user could only send messages after login with a OTP code
      • stored everything on mongodb, next.js only project, very simple
      • user could send feedback
    • we let users using the app for 7 days for free and we got a lot of feedback
    • from about 300 form subscribers we got about 70 of them to register to the app
    • most of users only send one message and never came back
    • some of them uses everyday from morning to night
    • we offered a stripe subscription for 20$ at the end of 7 days but got no sales
  • 2024/jul (first robust version): dochelp.ai
    • after the mvp test we got some insights
    • also decided to go into this as a co-founder partner
    • the team now was me (tech), marcio (doctor and main financer) and marcos (marketing and audio-visual)
    • i started developing this new version with one goal in mind: build a more robust, pretty and sellable version
  • 2025/feb: we launched!
    • i finally finished the app new version
    • we started to produce instagram ads again
    • with organic and free traffic we could get 100 users to subscribe
    • most of users don't use the app enough, most of them only send a test message and never come back
    • we noticed that maybe our target clients will only come with paid traffic
    • (on going) started planning sales workflow to focus on get some revenue

dochelp live website: https://dochelp.ai

wrapping up

if you're thinking about going solo as a technical founder, here's my advice: embrace the chaos, keep learning, and don't forget to take breaks (something i'm still working on!). it's a wild ride, but i wouldn't have it any other way.

catch you later!