
Learning Companion Agent
Use a working prototype to validate learning guidance, textbook tools, memory, and agent behavior limits
2025—present
Product experiments and independent builds beyond my shipped work — how I turn a question into a working prototype, then keep testing the product judgment and its technical limits.

Use a working prototype to validate learning guidance, textbook tools, memory, and agent behavior limits

Close handwritten OCR, sentence-level feedback, teacher review, and score validation into one traceable teaching workflow

First restore a multi-page paper into a reliable question structure, then generate marking, explanation, and knowledge-point diagnosis item by item

Turn affection for a game into a browser experience that can be published and kept improving

Organize long-term goals, real schedules, and execution feedback into a personal agent that can keep correcting itself