WABAC | Devpost

landing
dashboard
analytics

FOR EVALUATORS/SPONSORS:

Scroll down for a diagram-based guide to our architecture, key technologies used, and sponsor integration highlights.

💥 How it all started

42% of Americans aged 55 or older will eventually develop dementia, yet many digital solutions remain inaccessible or overwhelming for this demographic. We asked: What if support could be as simple as a phone call?

WABAC (pronounced “way-back”) is our answer—a memory-preserving voice AI assistant designed to help individuals living with dementia comfortably offload memories and receive cognitive support. By integrating Vapi's conversational voice agents and persistent memory tracking, we aim to provide both users and caregivers a tool for clarity, care, and connection.

📖 What it does

WABAC is a phone-based AI memory assistant that speaks directly with care receivers and automatically processes their input into meaningful insights for caregivers.

🧠 Core Features:

📆 Date-Stamped Memory Logging
Users can call and speak naturally. The assistant captures, time-stamps, and stores their thoughts and memories in a structured format.
📊 Caregiver Dashboard
Visualizes memory trends, reminders, and cognitive health metrics. Offers caregivers a transparent view of how their loved one is doing over time.
🔔 Smart Reminders
Events can be scheduled by either the caregiver or the user. The agent reminds users conversationally during future calls.
🧪 Cognitive Performance Monitoring
Tracks changes in memory consistency, vocabulary richness, and behavioral patterns to flag potential cognitive decline.

🔧 How we built it

Our architecture is simple but powerful—built around accessibility, persistence, and clarity. Below is a diagram of how data and interaction flow through our system:

WABAC System Diagram

🛠 System Overview

Vapi Voice AI Agents
- Handles natural phone-based conversations
- Allows care receivers to share memories and receive reminders
- Accepts inputs from caregivers via agent workflows
Database (Supabase)
- Stores memories, reminders, and cognitive metrics
- Indexed by user and timestamp for easy querying and trend tracking
Dashboard (Next.js + Chart.js)
- Visualizes memory history, reminders, and performance analytics
- Accessible to both caregivers and care receivers

🚩 Challenges we ran into

Natural data structuring: Extracting structured memory data from informal, voice-based input without making conversations feel robotic.
Voice UX: Making agent responses sound supportive while still capturing key details.
Data modeling: Designing a schema to track time, sentiment, and cognitive consistency longitudinally.

🏆 Accomplishments that we're proud of

Built an entire end-to-end system: voice → memory → dashboard
Successfully tracked real-time user memory input via phone and visualized it
Created a product that resonated emotionally with mentors and users alike

📝 What we learned

Nicholas:
I learned how to think deeply about user experience beyond UI—how something sounds matters when you're designing for care. Vapi and Letta allowed us to explore the balance between empathy and functionality.

Logan:
This was my first time working with conversational AI and persistent agent memory. I gained a lot of insight into system architecture and building for real-world users who may not be tech-savvy.