Learning Economy Foundation — AI-Powered Credential Generation

The Work

During my internship at Learning Economy Foundation, I built Boost Automation — an AI-powered tool that automated the creation of digital credentials using OpenAI's GPT and DALL-E 3 APIs.

Digital credentials are structured documents that certify a learner's achievements, skills, or qualifications. Before automation, creating these credentials required:

• Manual data entry for all credential fields (issuer name, description, criteria, etc.)
• Custom logo design for each credential
• Time-consuming formatting and validation

The goal: Use AI to generate structured credential data and logos automatically, reducing creation time from manual entry to seconds.

What I Built

AI Boost Wizard Interface

Designed and implemented the full user interface for the credential generation system using React and Tailwind CSS.

AI Boost Wizard category selection interface

The wizard guided users through:

1. Category Selection — Choose credential category (Work Skills, Life Skills, Soft Skills, Academics)
2. AI Generation — Use OpenAI to auto-generate credential fields
3. Logo Creation — Generate custom logos with DALL-E 3
4. Review & Finalize — Review and edit AI-generated content

Technical Implementation:

• Built multi-step form with React state management
• Created responsive UI with Tailwind CSS
• Implemented real-time validation and error handling
• Added loading states and progress indicators for API calls

OpenAI Function Calling Integration

Integrated OpenAI's function calling feature to generate structured credential data.

How it worked:

1. User provides basic input (credential name, category)
2. System constructs prompt with credential requirements
3. OpenAI GPT generates structured JSON with all credential fields
4. Function calling ensures output matches required schema

Technical Details:

// Define credential structure for function calling
{
  name: "generate_credential",
  description: "Generate structured credential data",
  parameters: {
    type: "object",
    properties: {
      name: { type: "string" },
      description: { type: "string" },
      criteria: { type: "string" },
      issuer: { type: "string" },
      skills: { type: "array" }
    }
  }
}

Console output showing OpenAI function calling response with structured credential data

Result: AI-generated credentials with properly formatted fields — issuer name, achievement description, earning criteria, associated skills, and more.

DALL-E 3 Logo Generation

Built integration with DALL-E 3 to automatically generate credential logos.

Implementation:

1. Construct image prompt based on credential category and name
2. Call DALL-E 3 API with specific style parameters
3. Retrieve generated image URL
4. Display preview and allow regeneration if needed

Prompt Engineering:

• Designed prompts to generate clean, professional logos
• Specified style constraints (flat design, simple shapes, educational context)
• Handled edge cases (inappropriate content, generation failures)

Example Prompts:

• "A simple, flat design logo for a Work Skills credential in technology, suitable for an educational certificate"
• "A minimalist icon representing Life Skills achievement, clean and professional"

Credential Preview & Editing

Credential form showing AI-generated fields for review and editing

Built the final review interface where users could:

• Preview AI-generated credential in badge format
• Edit any auto-generated fields
• Regenerate logo if needed
• Finalize and save credential

Technical Architecture

Frontend:

• React for component-based UI
• Tailwind CSS for responsive styling
• State management for multi-step wizard flow

AI Integration:

• OpenAI GPT-3.5/4 for structured text generation
• Function calling for schema validation
• DALL-E 3 for image generation

API Communication:

• Async/await for API calls
• Error handling and retry logic
• Loading states and user feedback

Key Technical Challenges

Challenge 1: Structured Output from LLMs

The Problem: LLMs generate free-form text, but credentials require specific fields in a structured format.

The Solution: Used OpenAI's function calling feature to enforce output schema. The model generates JSON that matches our credential structure, ensuring all required fields are present and properly formatted.

Challenge 2: Logo Generation Consistency

The Problem: DALL-E 3 can generate unpredictable results. We needed professional, appropriate logos every time.

The Solution: Carefully engineered prompts with style constraints. Added regeneration option for users. Implemented content filtering to catch inappropriate generations.

Challenge 3: User Experience During AI Calls

The Problem: API calls take 3-10 seconds. Users need feedback during processing.

The Solution: Built loading states with progress indicators. Added "Generating..." animations. Showed intermediate results when possible (e.g., field-by-field text generation).

Impact

What We Achieved:

• ✅ Automated credential creation: From manual entry to AI-powered generation
• ✅ Function calling integration: Structured data generation with schema validation
• ✅ DALL-E 3 integration: Automatic logo generation for credentials
• ✅ Full-stack implementation: React frontend with OpenAI backend integration

Time Savings: Reduced credential creation time from 15-20 minutes (manual data entry + logo design) to under 2 minutes (AI generation + review).

Technical Learnings

1. Prompt Engineering is Product Design

The quality of AI output depends entirely on prompt design. Small changes in prompt structure dramatically affected results. Learned to:

• Be specific about output format
• Provide examples in prompts
• Iterate based on actual generation quality

2. Function Calling for Structured Data

Function calling was game-changing for credential generation. Instead of parsing free-form text, we got structured JSON that directly mapped to our data model. Eliminated parsing errors and validation issues.

3. AI UX Requires Different Patterns

Traditional form UX doesn't work for AI generation. Users need:

• Clear loading states (AI is "thinking")
• Regeneration options (AI output isn't perfect)
• Edit capabilities (AI as starting point, not final output)

What I Would Improve

Short-term:

• Batch Generation: Generate multiple credentials at once
• Template Library: Pre-built credential templates for common use cases
• Version History: Track changes to AI-generated content

Long-term:

• Custom Training: Fine-tune models on existing credential corpus
• Multi-language Support: Generate credentials in multiple languages
• API Access: Let external systems integrate credential generation

Stack: React, JavaScript, OpenAI API (GPT + DALL-E 3), Tailwind CSS, Function Calling

Timeline: Summer 2023

Organization: Learning Economy Foundation (EdTech nonprofit focused on digital credentials and verifiable learning records)

Work: Boost Automation — AI-powered credential generation tool

Key Achievement: Reduced credential creation time from manual entry to automated AI generation using OpenAI function calling and DALL-E 3