BACKPERSONAL BUILDS
Personal Project
๐Ÿง 

Aphasia Toolkit

IN DEVELOPMENT

Speech Recovery ยท Cognitive Exercises ยท LLM-Powered Practice

Practice tools for aphasia recovery โ€” built for my father after his stroke. Not for a market. For a person.

Why I Built This

When my father had his stroke, I started looking for tools that could help him practice speech and cognition at home, between therapy sessions. What I found was dated, clinical software โ€” not built for someone sitting at a kitchen table trying to reconnect with language.

So I built my own. The goal wasn't to replace his therapists โ€” it was to give him something engaging he could use every day. Something that felt human, adapted to him, and made progress feel real.

What It Does

Speech Drills

Structured word and sentence repetition exercises that scale in difficulty. Tracks completion and response quality over time so progress is visible.

Photo-Based Q&A

A random photo is shown, and an LLM generates contextually relevant questions for the patient to answer verbally. No two sessions are the same.

Cognitive Exercises

Pattern recognition, word association, and memory tasks designed to supplement clinical therapy with daily lightweight practice.

Progress Tracking

Session history, completion rates, and response accuracy logged over time โ€” giving family members and caregivers visibility into day-to-day progress.

Adaptive Difficulty

The system monitors performance and adjusts exercise complexity up or down automatically, keeping the challenge in the right range.

Accessibility-First

Large text, high contrast, simplified navigation, and voice input โ€” designed for someone with motor and visual challenges, not just someone in a hurry.

Technical Approach

The core challenge was making LLM-powered features feel seamless on slow connections and with voice input that's often imprecise. The system uses a curated prompt pipeline to generate question sets from image analysis, then evaluates spoken answers with some tolerance for aphasia-related speech patterns.

Next.jsTypeScriptWeb Speech APIAnthropic APILLM IntegrationAccessibility (WCAG)
Active Development

This project is personal and ongoing. Features are added as my father's recovery evolves and his needs change. It's not a product โ€” it's a tool built for one person, with the kind of care that doesn't show up in a feature list.