Mvaak

The Mvaak Rehab App is a cutting-edge, AI-powered learning platform developed to support early-stage rehabilitation of children with dyslexia and other learning difficulties. Designed with a deep understanding of how neurodiverse children engage with language, the app offers a multisensory, adaptive, and regional language-first experience — making it one of the first tools of its kind tailored for the Indian education landscape.

At the core of the app lies a voice-first AI engine that facilitates reading support in both English and Kannada, allowing children to interact using their mother tongue. The platform integrates phonemic awareness activities (e.g., word vs. non-word recognition, syllable counting) and delivers real-time corrective feedback using speech recognition and audio event detection technology. By doing so, it helps children overcome decoding and pronunciation difficulties — foundational to reading fluency.

The learning journey is enhanced with:

• Gamified activities that keep children motivated
   
• Personalized learning paths based on each child’s pace and performance
   
• Visual, auditory, and tactile interactions designed for attention and retention
   
• Dashboards for parents and teachers to track progress, strengths, and areas for support

Key Differentiators

🎙️ Voice-First, AI-Powered Learning
Uses speech recognition and audio event detection to provide real-time, personalized feedback for dyslexic learners.

🌍 Regional Language Support (Kannada-first)
One of the few dyslexia tools designed for Indian languages, ensuring comfort and accessibility for mother-tongue learners.

🧠 Dyslexia-Specific Interventions
Activities tailored to core dyslexic challenges like phonemic decoding, syllable segmentation, and word recognition.

🛠 Adaptive Learning Engine
AI adjusts the difficulty and content flow based on each child’s pace, improving learning retention and engagement.

🏆 Gamified Learning Journey
Motivation-driven interface with rewards, progress maps, and playful interactions to sustain long-term engagement.

Resources required vary for the training and deployment scenarios 

• Training : 4-8 Nvidia A100 GPUs for model training and fine-tuning, 128GB RAM, 32-48 Core CPU, 1TB storage for datasets.
   
• Inference : 4-8 Core CPU, 4-8GB RAM, 256GB storage for model checkpoints and necessary libraries.