[Kwangwoon University 2025 Idea Commercialization Support Project] Melissa: An AI Diary App That Records Natural Conversations

Minseok Park, CEO of Melissa

Through text and voice conversations with a multi-agent AI, text and images are automatically generated.
It provides an easy and enjoyable diary-keeping experience without the burden of recording.

Melissa is an AI diary mobile app that automatically generates diaries and images through text and voice conversations with AI, helping users easily and enjoyably continue their own stories. It is currently available based on text conversations, and a voice conversation module is to be added within the year. It is available on both Android and iOS.

Melissa uses multi-agent AI to automatically generate summaries and images from text or voice conversations, making it easy and enjoyable for anyone to record their own stories.

Melissa’s core feature is interactive diary creation: simply conversing with the AI automatically records your day without needing to write anything. Based on a multi-agent architecture, each AI handles summarization, hash tag generation, and title/image creation to compile the user’s conversations into a complete diary. The diary is stored in a calendar-style interface for easy date-based review. Users can select or create their own AI persona, edit their diary as needed, or write multiple entries in a single day. Weekly and monthly reflections, slide shows, and push notifications help maintain consistent journaling habits. Completed entries or memorable conversations can be saved as images or shared on social media, enabling natural distribution of content. Technically, it combines a multi-agent architecture with a RAG-based memory system to deliver personalized responses. Later, this will expand into an immersive service through voice conversation and character-based interaction capabilities.

Melissa’s core competitive advantage lies not in its conversational AI ability, but in its structure that remembers and visualizes conversations, expanding into a sustainable record-keeping experience. While most AI services are limited to short conversation summaries, Melissa enhances both the immersiveness and persistence of records by combining hash tags, images, and titles to reconstruct the user’s day into a single narrative. Melissa combines a multi-agent system with a RAG-based memory architecture to deliver personalized conversational experiences, where users enjoy a natural journaling experience tailored to their unique conversational style and logging patterns. Through this differentiated approach, Melissa is evolving beyond a simple journaling app into a ‘journaling experience platform’ that users want to visit daily.

Technically, Melissa operates by integrating a multi-agent AI architecture with a memory system based on Retrieval-Augmented Generation (RAG). With the multi-agent AI architecture, users can directly create or select their desired AI persona. After each conversation, each agent simultaneously generates a conversation summary, title, hash tags, and images, which are assembled into a single, complete record. RAG is a user-specific memory system that stores and manages conversations and records as ‘memory chunks’. It provides responses customized to each individual’s conversation flow, generating increasingly natural and consistent dialogue over time. Conversations and diary summary data are stored encrypted using AES-256. It is protected from external intrusion by applying a TLS-based security protocol during data transmission. Currently operating on a text-based system, Melissa plans to introduce voice conversation and character-based interaction features in the future, expanding to an immersive recording experience that feels like conversing with a real person.

The Melissa team started with six members: Da-sol Kim (AI Architect), Min-seo Kim (UI/UX Designer), A-reum Yoo (Marketing Strategist), Hyun-sung Choi (Frontend Developer), Seong-hoon Jeong (Backend Developer), and CEO Minseok Park (Service Planning and Operations). Currently, three core members remain: CEO Minseok Park (Planning and Operations), Hyun-sung Choi (Frontend Developer), and Seong-hoon Jeong (Backend Developer). They collaborate with high dedication within their respective areas of expertise. CEO Park leads service planning and UX design, setting the overall direction. Hyun-sung Choi drives React-based interface development, app UI redesign, and performance optimization, enhancing the quality of the screens that users first encounter. Seong-hoon Jeong handles the entire infrastructure, including feature development, environment setup, AWS operations, API documentation, server security, and incident response. He is also in charge of Melissa’s backend design and stable system operations. The three members are organically collaborating under the shared goal of “solving the problem for people who understand the need for a diary but struggle to write consistently,” and are growing Melissa together.

Regarding future plans, CEO Park said, “We aim to release the final version of Melissa (1.3.0) by December this year,” adding, “We plan to enhance its completeness fully implementing key features, including the voice module, character dialogue, retrospective slide shows, push notifications, and diary writing/editing functions.”

Date of Establishment: January 2025
Main Business: Development and operation of the Melissa app service
Achievements: March 2025: Android and iOS app release; May 2025: Selected for Idea Commercialization Project; July 2025: Major update based on user feedback

reporter jinho lee
jinho2323@hankyung.com