Executive Summary – Voice-First Companion for Early Dementia

Inspiration We were inspired by the growing number of seniors in Singapore facing early dementia and the immense caregiver burden. Conversations with caregivers highlighted a gap: seniors need daily, gentle cognitive stimulation that is easy to access, culturally familiar, and respects their dignity.

Problem Singapore’s population aged 65+ is projected to reach 24.1% by 2030. Dementia affects ~8.8% of older adults (~73,900 in 2022), with over half undiagnosed. Caregivers experience high stress (23%–59%). Existing tools often fail due to low tech literacy, short attention spans, and limited engagement. Early intervention is critical: MCI → mild dementia stages are where cognitive exercises are most effective.

Evidence-Based Approach

Cognitive Stimulation Therapy (CST) and Reminiscence Therapy improve cognition, mood, and quality of life.

Spaced retrieval / repetition reinforces memory retention.

Short daily sessions maintain adherence and engagement.

Solution A Voice-First Memory Companion that provides:

Personalized prompts about life, family, and hobbies.

Spaced recall to reinforce memories.

Memory highlights in the user’s own voice for end-of-day playback.

Caregiver dashboard showing trends, adherence, and alerts.

Low-friction UX: wake-word or one-button capture, multilingual prompts, offline-first, and encrypted storage.

Development & What We Learned

Built using STT (Whisper) → embeddings → spaced repetition → similarity scoring → TTS playback → caregiver dashboard pipeline.

Challenges

Developing a usuable frontend for smoother experience.

Linking all different dependencies together to all work in a single build of python.

Building a usable database

Share this project:

Updates