Empathetic AI Diary

Project Introduction

• When users write a diary in the mobile application, it analyzes and categorizes it into one of six emotions using the KoBERT model
• A service that provides empathy and comfort tailored to user emotions using a fine-tuned LLM model
• Building a zero-downtime, high-availability service using AWS EKS and RDS

1. Technology Stack

1. AI and Models

   • KoBERT (Emotion Analysis)
   • LLM (Empathetic Dialogue Generation)
   
   

2. Cloud Infrastructure

   • AWS EKS (Kubernetes)
   • AWS RDS (Relational Database Service)
   
   

3. Development Environment

   • Python (KoBERT and LLM)
   • Container
   • React-Native Expo
   • Nest.js
   • PostgreDB
   • CoLAB
   
   

4. CI/CD

   • Jenkins
   • GitLab
   • ArgoCD
   • Harbor
   • SonarQube
   
   

5. Monitoring

   • Prometheus
   • Grafana
   • AWS CloudWatch
   
   

6. Collaboration Tools

   • Notion

   
   
   
   

2. Project Duration and Schedule Management

Project Duration

• September 13, 2024 ~ December 7, 2024

Schedule Management

• Task progress and resource management through Notion
• Weekly Agile scrum meetings

3. Project Differentiators

• Enhanced user experience through AI-based emotion analysis and personalized dialogue generation
• Providing stable service through high-availability infrastructure
• Improved accuracy and empathy in emotion analysis and response by combining KoBERT and LLM models

4. Improvement Goals

1. Model Optimization

   • Continuously improve the performance of KoBERT and LLM to provide better emotion analysis and empathetic dialogue
   
   

2. Enhance User Experience

   • Strengthen personalized services through diverse emotion analysis and empathetic responses
   
   

3. Strengthen Infrastructure Scalability

   • Continuously optimize high-availability infrastructure to prepare for an increase in the number of users