AdaptIQ

• 

• 

• 

• 

• 

🚀 Inspiration

Children with disabilities often face barriers in accessing mainstream education due to a lack of adaptive learning resources. Whether it's a blind student struggling with text-based materials, a deaf or mute child finding it hard to communicate in class, or neurodivergent learners like those with autism or ADHD needing personalized guidance — the existing systems fail to accommodate their needs effectively. Inspired by the belief that education should be universally accessible, we built AdaptIQ, an inclusive AI-powered platform designed to bridge the accessibility gap.

🧠 What it does

AdaptIQ is an assistive, browser-based platform for children with diverse learning and communication disabilities. It offers:

🎙️ Sign-to-Voice & Voice-to-Sign Interpreter: Converts live speech into sign language animations and vice versa using AI and video input.

📚 Adaptive Learning Assistant: Learns from the user’s pace and behavior to deliver personalized educational content based on their disability (blind, deaf, ADHD, autistic, etc.).

📖 Text-to-Speech & Speech-to-Text: Converts reading material into speech for blind students and transcribes speech into text or sign for the hearing impaired.

🔒 Offline + Privacy-First Mode: No sensitive data leaves the device. Everything runs locally using open-source LLMs.

🛠️ How we built it

Frontend:

TypeScript + Next.js (App Router)

Tailwind CSS with Radix UI + ShadCN UI for accessibility-friendly components

Webcam & microphone integration for sign/speech input

Backend / AI Layer:

Ollama LLM – lightweight, locally running open-source language model

Whisper.cpp (or alternates like DeepSpeech) for offline STT

TensorFlow.js for sign recognition & animation

Supabase for optional auth & session tracking

DevOps & Hosting:

Hosted on Vercel

GitHub Actions for CI/CD

🧩 Challenges we ran into

Getting reliable sign language translation required dataset tuning and real-time webcam optimization.

Whisper API was paid, so we had to experiment with open-source STT models compatible with JavaScript.

Adapting LLMs like Ollama to run locally in-browser (via WASM or edge functions) with privacy was tricky.

Accessibility testing for each disability required simulated environments and test cases.

🏆 Accomplishments that we're proud of

Successfully enabled bidirectional sign-speech translation with real-time feedback.

Integrated a customizable learning assistant that adapts based on disability profile.

Built a fully offline-compatible, privacy-first educational platform.

📚 What we learned

How to use Ollama LLMs effectively without relying on GPT APIs.

Ways to integrate accessibility-first UI principles into design using Radix + ShadCN.

The limitations and workarounds for browser-based AI models (especially for sign recognition).

The importance of inclusive design thinking when creating for underserved users.

🔮 What's next for AdaptIQ

Adding support for regional sign languages (e.g., Indian Sign Language).

Integrating emotion detection to help with autism spectrum support.

Deploying a PWA version for offline mobile use in rural/low-connectivity areas.

Collaborating with special educators and therapists to refine our modules.

Built With

• css3
• elevenlabs
• json
• next
• ollama
• react
• typescript