AI for Transparent Elections – In Person Hackathon

AI for Transparent Elections – Hackathon Join us to build practical AI tools that support transparency and quality control in election processes. 28th of March, Barter Community Hub, bul. Cherni Vrah 47, Rooftop, 1407 Sofia What is it? AI for Transparent Elections is a one‑day civic technology hackathon focused on building practical AI applications that improve transparency and quality control in election processes. Participants will work with real datasets from previous elections — including scanned paper protocols, electronic protocol records, and video streams from vote counting procedures. The goal of the event is not only to experiment with machine learning models, but to build working applications that can run online and automatically analyze election data. These tools should help observers, analysts, and researchers detect inconsistencies faster and monitor potential issues at scale. Participants will collaborate in small teams to prototype systems that analyze handwritten election protocols, monitor video streams from vote counting environments, and detect statistical anomalies across multiple data sources. Organized by AI Engineer Foundation Europe, Data Science Society and AI activists like Victor. Goals Teams will work on building applications addressing three main goals. I. Election Protocol Analysis Build systems that can: Automatically analyze handwritten election protocolsDetect corrections, overwritten numbers, and inconsistenciesIdentify arithmetic anomalies in protocol dataExtract structured data from scanned protocolsCompare extracted values with electronic protocol recordsHighlight potential mismatches between paper and digital data II. Video Monitoring Alert System Develop systems that can: Monitor video streams from vote counting environmentsDetect missing or interrupted streamsIdentify low‑quality audioDetect lack of visible activityIdentify inadequate camera positioningFlag technical or procedural issues in the monitoring process III. Statistical Analysis at Scale Develop tools that combine multiple sources of information to flag potential risks in election data by analyzing: scanned paper protocolselectronic protocol recordsvideo monitoring streamspublic election predictions or reference datasets The goal is to detect unusual patterns or inconsistencies across large datasets and provide signals that may require human review. Datasets Participants will work with prepared datasets based on publicly available data from previous elections. The datasets include: Election Protocol Data Scanned paper protocolsExamples with handwritten valuesProtocols containing corrections and overwritten valuesProtocols with known arithmetic inconsistenciesCorresponding electronic protocol records Video Monitoring Data Video recordings from vote counting proceduresExamples with different camera anglesVarying audio and video qualityCases with technical interruptions or incomplete visibility These datasets allow teams to work with realistic scenarios and develop solutions that address real-world challenges. Format The hackathon will take place on March 28. Participants will form several teams of 5-6 people each and work collaboratively throughout the day to design and build prototype applications. Schedule 09:00 - 09:30Registration 09:30 – 10:30Introduction to the problems, presentation of the datasets and testing environment, and team formation. 10:30 – 12:30Coding session. 12:30 – 13:30Lunch break. 13:30 – 16:30Coding session. 16:30 – 17:00Discussion over coffee or beer. 17:00 – 19:00Coding session. 19:00+Short presentation of results, discussion, and preparation of deliverables. 21:00End of event. Deliverables At the end of the hackathon each team is expected to produce: Working prototype applicationsTools capable of processing individual protocols and large sets of protocolsSystems capable of analyzing video streams for monitoring issuesSource code ready for open‑source publication on GitHubShort demo videos showing how the applications workBasic documentation describing the approach and architecture The goal is to produce practical prototypes that can be further developed into open tools supporting election transparency. Who Should Join We welcome participants with backgrounds in: Machine Learning / AIData ScienceComputer VisionSoftware EngineeringData EngineeringCivic TechnologyUX / Data Visualization Interest in election transparency, data analysis, and collaborative problem solving is highly encouraged. Join Us If you are interested in applying AI to real‑world civic challenges and collaborating with other engineers and researchers, we would love to have you join us.