Inspiration
As wikimedia editors the is an obligation to ensure that everyone is represented. Some communities have also realized this misrepresentation so they created initiatives like Wiki Loves Women to lift the voices of people not heard. This tool is a way of supporting this initiatives to ensure the impact is more meaningful and easier to archive .
What it does
Scans Wikipedia/Wikidata datasets. Uses Gemini 3 reasoning to detect patterns of bias and missing Generates actionable edit suggestions (e.g., “Add biography of X, missing despite reliable sources”). Explains why gaps exist (e.g., systemic bias, lack of citations, language barriers). Horizon is an AI-powered platform that detects, quantifies, and explains representation gaps in public knowledge. Using structured data from Wikidata, Horizon provides a clear overview of who and what is missing from global knowledge systems—across gender, geography, language, and topic areas.
The platform separates measurement from reasoning: an Overview dashboard presents transparent statistics through charts and metrics, while an Analyse section uses Gemini 3 to interpret these statistics, explain why gaps exist, and recommend high-impact actions. Rather than showing static numbers, Horizon transforms data into insights that communities, educators, and researchers can act on to improve knowledge equity.
How we built it
Horizon is built as a full-stack application. We use Wikidata as a structured proxy for Wikipedia coverage, querying it to generate verifiable statistics on biographies, topics, and representation. A FastAPI backend aggregates this data and exposes clean, chart-ready APIs.
We integrate Gemini 3 in a dedicated reasoning layer, where it analyzes the factual statistics to identify critical gaps, explain systemic causes, and generate actionable recommendations. The frontend is built with Next.js, featuring a modular dashboard with stat cards, pie charts, bar charts, and AI-generated insight panels. This architecture ensures transparency, scalability, and clear separation between data and AI reasoning.
Challenges we ran into
Due to limited funds, we hit the rate limit very quickly during testing so we go two api keys each to speed up development. hallenge was designing the system so that AI did not generate or hallucinate statistics. We addressed this by strictly separating numeric computation (Wikidata + backend) from interpretation (Gemini 3). Another challenge was modeling “underrepresentation” in a way that is measurable yet meaningful across countries and topics. We also had to carefully design APIs and frontend state management to ensure the Overview and Analyse sections stayed consistent and linked to the same source of truth.
Accomplishments that we're proud of
We are proud of building a non-chat, AI-native analytics platform that goes beyond surface-level dashboards. Horizon doesn’t just visualize gaps—it reasons about them. The clear separation between overview metrics and AI analysis makes the platform trustworthy, explainable, and extensible. As a student-built project, delivering a complete end-to-end system that combines data engineering, API design, AI reasoning, and frontend visualization is a major achievement for our team.
What we learned
Bring an idea to reality is very difficult especially if they are beyond your comfort zone. We learned how to design responsible AI systems where models explain data rather than invent it. We gained hands-on experience integrating large language models into real applications, structuring APIs for analytics dashboards, and working with large open datasets like Wikidata. Most importantly, we learned how thoughtful system design can turn raw data into meaningful, human-centered insights.
What's next for Horizon
Next, we plan to expand Horizon to support cross-country comparisons, language-specific analysis, and article quality signals. We also aim to add collaboration features that allow communities to track progress over time and measure the impact of interventions. In the long term, Horizon could become a global knowledge equity observatory—helping institutions, educators, and contributors make informed decisions about where to focus their efforts.
Log in or sign up for Devpost to join the conversation.