← Back to Portfolio
Personal · project

Perfect Prompt: an AI image-prompt optimizer

Award-winning Streamlit application that won first place in Cohere Hackathon #3 2022, leveraging ML classification to automatically identify artistic styles and generate optimized prompts for image generation workflows

A hackathon-winning application that solves the complex challenge of prompt optimization for AI image generation. Perfect Prompt combines machine learning classification with automated prompt enhancement to help users create more effective prompts across six distinct artistic styles including cyberpunk, steampunk, watercolors, cottagecore, landscapes, and architecture. The solution streamlines the traditionally iterative process of prompt engineering into an automated workflow that delivers optimized variations in under a minute.

Key Features

Intelligent Style Classification: Automatically identifies artistic style from user input using Cohere's classification API across six predefined categories • Automated Prompt Generation: Creates multiple optimized prompt variations tailored to the identified artistic style for enhanced image generation results
Real-time Processing: Delivers prompt optimization and style matching in under 60 seconds through efficient API integration • Production-Ready Deployment: Features live Streamlit deployment with integrated secrets management for secure API key handling

Technical Implementation

The architecture demonstrates sophisticated integration of multiple AI APIs, combining Cohere's natural language classification capabilities with Replicate's image generation services through a clean Streamlit interface. The solution showcases practical machine learning application design, implementing secure API management patterns and efficient data processing workflows that bridge natural language understanding with creative AI tools.

Skills & Technologies

Python Streamlit Cohere API Machine Learning Classification Replicate API Pandas AI/ML Integration Web Application Development
// stay in touch

Get Answering Machines in your inbox.

Occasional notes on AI, coding agents, and where this is all headed — plus the very occasional origami diagram.

Arjun Kirti Patel
Chicago, IL · folded with care
© 2026 Arjun Kirti Patel · MIT Licensed