DEV Community

local ai
local ai

Posted on

How I Built an AI Tool for Scientists and Grew to 10K Users as a Solo Founder

Hey dev community! πŸ‘‹

I want to share my journey building a niche AI SaaS product as a solo founder. Hopefully, some insights here will be helpful for others on a similar path.

The Problem I Noticed

While working with researchers, I noticed a recurring pain point: scientists spend hours creating figures for their papers. Most use PowerPoint or struggle with Illustrator, even though they're not designers. The results often look unprofessional, and the process is frustrating.

The Solution

I built SciDraw - an AI-powered platform specifically for scientific illustrations. Researchers can describe what they need or upload a rough sketch, and the AI generates publication-ready figures.

Key features:

  • Text-to-image generation for scientific concepts
  • Sketch refinement (turn rough drawings into polished diagrams)
  • Editable SVG export (so users can fine-tune in any vector editor)

Tech Stack

For those curious about the technical side:

  • Frontend: Next.js + React
  • Backend: Node.js
  • AI: Custom pipeline combining multiple image generation models
  • SVG Processing: Custom algorithms to ensure clean, editable vector output

What Worked for Growth

I didn't have a marketing budget, so I focused on:

  1. SEO - Targeting long-tail keywords like "AI scientific figure generator"
  2. Directory submissions - Listed on 80+ platforms (Product Hunt, G2, etc.)
  3. Niche focus - Instead of competing with general AI image tools, I went deep into one vertical

Current Stats

  • 10,000+ researchers using the platform
  • ~$2K MRR
  • Users from Stanford, MIT, Harvard, and universities worldwide
  • Fully bootstrapped, no funding

Lessons Learned

  1. Niche down aggressively - A small pond is easier to dominate
  2. Solve a real pain point - Scientists were literally wasting hours; the value prop was obvious
  3. Talk to users - Early feedback shaped the product significantly
  4. SEO compounds - It's slow at first, but worth the investment

What's Next

Currently working on:

  • More diagram types (molecular structures, experimental workflows)
  • Collaboration features for research teams
  • API for integration with other tools

Happy to answer any questions about building for the academic market or AI image generation!

What niche are you building for? πŸ‘‡

Top comments (0)