DataSprint 3.0

DataSprint 3.0 is a 24-hour offline Data Science Hackathon preceded by an online preliminary screening round. This event is designed to evaluate how participants think with data, not just how they code. Participants will work with realistic, noisy logistics datasets and solve progressively challenging analytical problems that reflect how data engineers and data scientists operate in real organizations — dealing with messy inputs, hidden signals, and the need to draw responsible, explainable conclusions under time pressure. The hackathon emphasizes: Analytical reasoning over raw model accuracy Feature engineering from real-world-like data Data cleaning and visualization for insight generation Logical conclusions from imperfect information Clear data storytelling backed by evidence Join the Whatsapp channel: Follow the Data Sprint 3.0 channel on WhatsApp: https://whatsapp.com/channel/0029VbCR43k9hXFFUhOHpB2E Event Structure Round 1 — Prelims (Online Screening) This round is conducted to shortlist teams for the offline hackathon. Mode: Online Participants solve analytical and logical questions based on a sample dataset Focus on data interpretation, reasoning, and basic ML understanding Duration: 30 mins Shortlisted teams qualify for the 24-hour offline hackathon Round 2 — 24-Hour Offline Hackathon Shortlisted teams participate in the main DataSprint 3.0 experience. During the hackathon, teams will move through multiple progressive data investigation stages, where each stage reveals a new operational problem hidden within the system such as: Delivery delay analysis SLA violation investigation Trend and performance drift detection Automation decision risk analysis In each stage, teams must: Clean and analyze raw datasets Build dashboards and visual insights Perform feature engineering Apply ML models (classification and regression) Answer investigative questions through a submission portal Perform prediction tasks on given sample inputs Team Formation: Team size: must contain 4 members Inter-college teams are not allowed Inter-specialization teams are allowed Individual participation is not allowed Eligibility: Open to all UG students Basic knowledge of Python, Data Analysis, Visualization, and ML is recommended Participants who are shortlisted for a offline hackthon must bring their own laptops, ethernet cables with required software Tools Allowed: Python (Pandas, NumPy, Scikit-learn, etc.) Tableau / Power BI  Jupyter Notebook / VS Code Rules of the Hackathon: All work must be done within the hackathon duration. Pre-built models, notebooks, or prepared solutions are not allowed. Plagiarism or unfair means will lead to disqualification. Answers must be submitted only through the provided portal. Teams may be asked to explain their approach to judges. Judges’ decisions are final. Evaluation Criteria: Teams will be evaluated based on Ability to identify true drivers hidden in data Quality of Dataset after preprocessing Distinguishing misleading signals from meaningful patterns Quality of analysis, dashboards, and insights Correctness of predictions Logical reasoning and justification of answers