The second edition of Hacking Big Numbers (HBN2) has successfully concluded in Bucharest, marking a significant milestone in the intersection of artificial intelligence and business data analysis.
This innovative hackathon, centered around the theme "Solving Big Real-World Problems with AI and Business Data," brought together some of the brightest minds in data science, machine learning, and software development.
The event's centerpiece was an intense Machine Learning (ML) tournament that challenged participants to push the boundaries of industry classification accuracy using the North American Industry Classification System (NAICS).
This unique competition format, reminiscent of a high-stakes game of '20 Questions,' tested participants' abilities to develop sophisticated AI models capable of identifying a company's core business with minimal information.
The tournament's innovative format challenged competitors through a series of five-round “head-to-head” games, each focusing on a single company.
Starting with just a commercial name, participants received increasingly detailed information in subsequent rounds, including business tags, short descriptions, full company profiles, and finally, the company's category.
This gradual reveal of data tested the models' ability to make accurate predictions with limited information and adapt quickly as more details became available. Under strict time constraints of just 5 seconds per prediction, participants applied their models to classify each company's industry.
The scoring system added another layer of complexity, heavily rewarding early correct guesses while increasing penalties for mistakes in later rounds. This structure encouraged the development of models capable of making accurate predictions with minimal data, mirroring real-world classification challenges.
The tournament not only showcased the analytical capabilities of the participants but also demonstrated the transformative potential of AI in industry classification. The winning team achieved an impressive 70% accuracy rate across the selected dataset, with a solution developed in jut one weekend.
For context, Veridion, a leader in the industry, boasts an accuracy rate of over 90% for 115M+ company profiles, setting a high benchmark for future competitions.
Supported by industry leaders such as Banca Transilvania and Underline Ventures, HBN2 offered a diverse range of challenges beyond the ML tournament. These included developing an Anti-Money Laundering (AML) detection system, creating AI-driven competitor analysis tools for startups, and enhancing data integrity through advanced analytics.
Companies with valuable internal data interested in proposing a challenge for the next edition can contact Veridion at cosmin.pirvu@veridion.com to explore partnership opportunities for HBN3 and drive innovation in their industry.
Participants at the hackathon were provided with access to an extensive array of resources, including Veridion's comprehensive datasets, dedicated servers for model training, and exclusive data from commercial registers and social media platforms. These tools enabled them to harness the full potential of AI technologies and contribute actionable insights into various business domains.
Veridion is a next-generation data provider specializing in company data, maintaining a weekly updated database of over 115 million company profiles.
Their data covers every aspect of a business, from identification, locations and classification to product offerings, ESG topics and more.
Veridion uses advanced scraping and machine learning technology to understand the ever-expanding business landscape and deliver the resulting data through APIs tailored to data science teams, enabling business decision-making at scale.
In addition to the hackathon, Veridion offers an exclusive Startup Program designed to empower early-stage companies with access to high-quality business data at a significantly reduced cost.
Startups can leverage Veridion’s Match & Enrich API to automatically augment their databases with detailed company profiles, and the Complex Search API to identify potential new clients, investment opportunities, or markets.
The program offers a 70% discount compared to regular pricing, with plans starting as low as $50 per month. All innovative startups are encouraged to join the startup program and unlock the full potential of their data-driven strategies.
For more details and to sign up, visit the dedicated website (https://veridion.com/startup-program/) or contact Cosmin Pirvu, Startup Program Manager at Veridion, at cosmin.pirvu@veridion.com.