ClassSnap

Overview

A reading platform that builds English comprehension in deliberate steps rather than serving random passages. Built as co-founder and CTO of EnglishAI, with ownership of product architecture and implementation.

Core Problem: Generated Text Is Easy; Pedagogically Ordered Text Is Not

An LLM will produce an endless supply of English passages — which is not the same as a curriculum. Reading ability develops through a sequence: specific skills, in a specific order, at a difficulty held just above the learner. Text generated without those constraints produces motion without progress.

The generation engine was therefore built to answer to pedagogy rather than the reverse. The target skill and the learner’s current level act as inputs that constrain each passage, and difficulty adapts from measured performance rather than self-reported level.

Key Contributions

• Designed the reading-skill learning system on established pedagogical theory, so the sequence of skills — not the content pipeline — drives the product
• Built the adaptive passage generation and recommendation engine on Gemini, with difficulty bounded by the learner’s measured level
• Developed progress tracking and personalized feedback — converting per-question results into a form a teacher can act on
• Designed and implemented the teacher-student dashboard — reflecting that classroom adoption hinges on the teacher’s view, not the learner’s

Tech Stack

• AI: Google Gemini
• TTS: Azure Speech
• Frontend: Next.js, React, TypeScript
• Backend: Supabase
• UI: Tailwind CSS, Radix UI, Lucide React
• Infrastructure: Vercel