StockEcho
Period: 2026
Code: github.com/kimminjunnn/StockEcho
Overview
StockEcho organizes news and corporate disclosures about the stocks a user holds into events. It then shows how prices reacted to similar past events and how those reactions would affect the user’s portfolio weights. It is an educational analysis tool: it explains historical scenarios with sources and does not predict future returns.
My role
- Planning: Defined user questions and the core flow; designed requirements for Risk Replay, portfolio-impact views, and evidence display.
- Back-end and data: Connected KIS price data (current and daily), NAVER News, and OpenDART disclosures. Designed relevance classification, de-duplication, trading-day return calculation, and caching, fallback, and security policies.
- Front-end: Helped implement stock selection and weight validation; analysis-result, news, disclosure, and price-reaction screens; and loading, error, and stale-data states.
Design principles
- Calculation and explanation are separated: numbers come from the analysis engine, and the LLM only writes evidence-based explanations.
- Every key claim links to its source (title, publication date, original link).
- Sample size, similarity, data timestamp, and limitations are shown with each result.
