AWS Builder Center

Weekend Creative Challenge: CoreCheck.exe

an unhinged ai fashion roast machine that grades your fit, measures your vibe, and tells you if you ate or are completely cooked.

Vision & What the App Does

CoreCheck.exe is an unhinged, AI-powered aesthetic and fashion vibe grader designed to tell you the raw truth about your outfit. Whether you are aiming for cottagecore, clean girl, dark academia, mob wife, gorpcore, or y2k, CoreCheck takes your outfit photo, analyzes every visual detail through the multimodal capabilities of Gemini 3.5 Flash, and delivers a brutal yet hilarious Gen-Z critique scorecard.
The creative output includes three core elements:
  1. The Match Percentage: An exact score from 0 to 100% measuring how close your fit is to your chosen aesthetic.
  1. The Verdict: A razor-sharp, custom roast sentence written in authentic internet slang (evaluating whether you "ate," served a "main character aura," or are completely "cooked").
  1. The Gap List: A tactical checklist of three specific accessories or style items missing from your look that would instantly raise your score.
By turning fashion analysis into a gamified, humorous experience, CoreCheck provides instant entertainment while helping users experiment with new personal styles.

How You Built It

Building CoreCheck.exe was a masterclass in rapid prototyping, architectural pivoting, and serverless engineering over a single weekend.

Development Process & Key Decisions

The application is split into a sleek, retro-futuristic terminal UI on the frontend and a lightning-fast serverless backend driven by high-speed multimodal intelligence.
  • Frontend: Built using React, Tailwind CSS, and Lucide icons, styled with a high-contrast dark-mode terminal aesthetic. It handles image uploads, base64 conversion, aesthetic selection, and renders live scorecards alongside session progress history.
  • Backend: Built using Python on AWS Lambda, exposed publicly via Amazon API Gateway HTTP APIs, backed by Amazon DynamoDB for persistent session tracking, and integrating the Gemini 3.5 Flash API to process image bytes and generate structured JSON critiques at lightning speed.

Challenges Encountered & How I Overcame Them

The biggest hurdle I encountered was navigating initial model access bottlenecks and sandbox constraints on cloud-native services. When standard sandbox limits proved restrictive, I pivoted strategically: I kept my entire compute, routing, and database layer 100% native to AWS, while securely orchestrating vision requests through an optimized serverless pipeline calling Gemini 3.5 Flash.
Additionally, deploying the frontend revealed classic production pitfalls, such as Vite environment variable bundling issues and browser CORS preflight ( OPTIONS ) blocks. I solved this by explicitly configuring HTTP API Gateway CORS headers, handling preflight methods natively in Lambda, and locking down clean API endpoint communication so production builds connect seamlessly without latency or fallback loops.

AWS Services Used / Architecture Overview

CoreCheck.exe relies entirely on a modern, scalable, serverless architecture that scales down to zero cost when idle and handles high-concurrency requests instantly.

AWS Services Utilized

  • Amazon API Gateway (HTTP API): Acts as the secure, low-latency entry point for all frontend requests, routing  /score ,  /cores , and  /history  endpoints.
  • AWS Lambda (Python 3.12): Executes core business logic, parses base64 images, coordinates Gemini 3.5 Flash multimodal vision prompts, and formats structured JSON responses.
  • Amazon DynamoDB: A NoSQL table storing user session data, tracking previous style attempts, and calculating score deltas over time ( match_pct  improvements across iterations).
  • Amazon S3 / Amplify: Hosts the high-performance static frontend application assets.

Architecture Flow Diagram

[ User ]
│ (HTTPS POST / JSON + Base64 Image)
▼
[ Amazon API Gateway (HTTP API) ]
│
▼
[ AWS Lambda (Python Backend) ] ◄───► [ Amazon DynamoDB ]
│ (Session & History Tracking)
│ (Multimodal Prompt & Image Base64)
▼
[ Gemini 3.5 Flash API ]


What You Learned

This weekend challenge pushed my limits and yielded several valuable technical takeaways:
  • Multimodal AI Integration: Gained hands-on experience structuring prompts and passing high-resolution base64 image data to Gemini 3.5 Flash to return reliable, structured JSON critiques.
  • Serverless Mastery: Gained deep experience configuring stateless AWS Lambda handlers and routing lightweight endpoints via Amazon API Gateway HTTP APIs.
  • CORS & Production Resilience: Learned how critical proper preflight ( OPTIONS ) handling and strict header configuration are when bridging serverless backends with modern single-page frontend frameworks in production.
  • State Management via NoSQL: Leveraged DynamoDB partition keys to track iterative user progress seamlessly, allowing users to measure their fashion improvement journey over time.

Link to App or Repo

Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article