Aprotly - AI prompt engineer

Create context-rich AI prompts simply by chatting, not engineering

Comment

Inspiration

AI is everywhere, and everyone wants high-quality results with minimal effort. The problem is simple: AI without context produces bad output. Most users don’t know how to write good prompts, so even powerful models fail to understand what they want.

I personally faced this issue while generating images using Nano Banana. Gemini couldn’t understand my request because my prompt was vague and poorly structured. That’s when it became obvious: the problem isn’t the AI, it’s the prompt.
So I decided to build an app that helps people generate high-quality prompts without needing prompt-engineering skills.

What It Does

The app guides users through prompt creation by asking structured questions first, such as:

  • What do you want to generate?
  • What style or colors should be used?
  • Any specific constraints or preferences?

Users answer these questions, and the AI converts the responses into a clear, context-rich prompt.
Simple rule: better answers → better prompts → better AI results.

How We Built It

We built the project quickly by focusing only on the core feature and intentionally skipping authentication and databases.

Tech stack:

  • Next.js for development
  • TailwindCSS for styling
  • shadcn/ui for UI components
  • Gemini (Google AI Studio) as the core AI model

We experimented with multiple Gemini models, including:

  • gemini-2.5-flash
  • gemini-pro
  • other variants for performance testing

Gemini handles both the questioning flow and final prompt generation.

Challenges We Ran Into

This project looked simple at first. It wasn’t.

  • We ran out of Gemini credits during testing
  • AI responses were slow due to large system prompts
  • We lacked strong design experience
  • ElevenLabs integration failed and our API access got disabled

We solved these by reducing prompt size, caching system prompts, optimizing requests, and simplifying the UI until the app became stable.

Accomplishments That We’re Proud Of

  • Fixed major performance issues
  • Stabilized a broken AI pipeline
  • Turned a weak design into a clean, pixel-perfect UI
  • Shipped a working product under pressure

Each of these required real debugging and persistence.

What We Learned

  • Long system prompts are expensive → caching saves credits and time
  • Design improves through iteration, not talent
  • Consistency matters more than motivation

What’s Next for Aprotly – AI Prompt Engineer

Right now, the product focuses only on core functionality.

Planned improvements:

  • Authentication
  • Database for saved prompts
  • “Speak to improve prompt” feature
  • Better personalization and prompt history

This is only the foundation. The real work starts next.

Built With

Share this project:

Updates