⛅A personalized weather app for user's health⛅
👉 "LLM-based weather assistant app" (final concept developed through the course)
🎩 Members
기나연 (20203576) - PM
유저 페르소나 및 플로우 설계
프론트엔드 UI 개발 및 전체 디자인
기온·미세먼지 그래프 애니메이션 구현
Firebase 기반 사용자 프로필 전환 기능
LLM 응답 다국어 자연어 질문 처리 및 개인화 보완
김도연 (20220462)
위치 기반 API 관련 기능
기온·미세먼지 그래프 시각화
LLM 아키텍처 재구성
Git 관리 및 Tech Lead
이하형 (20211591)
LLM 개인화 응답 생성
자연어 입력에 대한 정보 추출
Firebase 연동, 사용자 데이터 저장 및 로드
홍순일 (20213592)
미세먼지/꽃가루 API 관련 기능
음성 인식 기능 구현
앱 빌드 및 배포 담당
🔍 Problems We Need to Solve
Main Problem
Users have difficulty quickly identifying weather information that affects their health and responding appropriately.
Sub-Problems
Weather changes can trigger various health issues such as allergies, asthma, headaches, and skin irritation. However, many people are unaware of how the weather affects their health, or they don't receive timely information, which delays appropriate responses.
People with different health conditions are affected by different weather elements. For example, people with pollen allergies are sensitive to pollen levels, people with sensitive skin are sensitive to UV exposure, and people with respiratory conditions are sensitive to fine dust levels.
Sensitivity to different weather elements also varies by age. Elderly people are more vulnerable to changes in temperature or atmospheric pressure, which can lead to symptoms such as low blood pressure, dizziness, or joint pain. In contrast, children are more vulnerable to pollen, fine dust, and UV rays, which can lead to respiratory issues or skin irritation.
Most weather apps don't consider people's different health needs or concerns. They just provide the same general weather information to everyone.
Many users feel frustrated because weather apps show a lot of information they don't really need, and it's hard to quickly find what they're looking for. So in the end, they often search things like "Seoul weather" or "Seoul fine dust" on the web instead, which feels inconvenient.
Most weather apps only provide simple alerts such as "Poor air quality" or "High pollen levels" without offering practical advice on what users should do. For example, clear guidance like "Today's pollen level is high. Please rinse your nose after going outside and consider taking allergy medication in advance" is usually not included.
🎯 Our App's Purpose
To provide personalized weather information based on each user's health needs or concerns, helping them take timely actions to protect their health.
👥 Our App's Audience
Health-conscious individuals: People who care about how weather affects their health and want relevant information for prevention or self-care
People with weather-sensitive health conditions (e.g. allergies, asthma, sensitive skin, migraines, etc.)
Users who want personalized weather insights: People who prefer to filter out unnecessary data and only see weather elements that matter to them. (e.g. UV index, pollen, air quality, etc.)
People of all ages (e.g. parents of young children, students, working adults, and elderly people)
🕵️♀️ Research the competition
🔍 In-depth Interviews
🔗 https://docs.google.com/spreadsheets/d/115pAeKiFazHTj4nsRoBl4nrImfUl-xizCywbzVYETTI/edit?usp=sharing
We conducted in-depth interviews with 15 participants to better understand users’ perspectives.
🧩 Thematic Coding
⭐3rd-Level Codes(Core Features)⭐
LLM-Based Conversational Weather Search: Users can ask weather-related questions in natural language via voice or text. The LLM interprets these queries and fetches relevant data from weather APIs to generate personalized responses. (➡️ Among various LLMs, we chose Google's Gemini for this project.)
Personalized Weather Data Visualization: Based on the user's question, only the most relevant weather information is visualized through clear graphs and intuitive visuals, making it easy to understand at a glance.
Weather Response Guide: Depending on the weather conditions, the LLM generates personalized guidance in natural language—such as what to bring, how to act, and what health precautions to take.
👩🏻💼 Persona
✏️ Wireframes and Mock ups