Built to meet the exacting standards of Winton's data division, Hivemind has for several years provided Winton researchers with a unique capability for industrial-scale data collection from primary sources.
Now available to other financial institutions and data-driven organisations, Hivemind provides a secure cloud-based infrastructure for the design and management of data projects. The platform allows for complex problems to be broken down into a hierarchy of atomised tasks, which are discrete enough to be completed with AI tools or to be distributed to a human workforce. Integrated consensus-based data quality assurance vouches for the integrity of the final results.
Hivemind's data scientists have extensive experience of breaking down and structuring otherwise unfeasible projects via its platform. In the last 12 months, Hivemind has completed nearly 2.5 million tasks, utilised 90,000 hours of human input, and produced around 20 proprietary datasets for Winton's research group. This team is on call to provide guidance, expertise and consultancy services over the life-cycle of client projects.
Example applications include:
- Creating structured data sets from unstructured sources (e.g. newspaper archives, regulatory filings, images and videos, news feeds, time series plots and multi-lingual content)
- Image / video / text categorisation
- Creation of Machine Learning training data
- Parsing and extraction from financial reports
- News sentiment analysis
- Customer database validation
- Cleaning market data feeds