Kipu Quantum, a leading provider of quantum software applications, announced general availability of Rimay Quantum Feature Extraction, a service proven to boost the performance of classical machine learning (ML) models. Rimay integrates into existing ML pipelines and enhances model accuracy by extracting richer quantum features from the same data. This applies particularly to cases where data is scarce, noisy, or imbalanced.
Enterprise users across manufacturing, financial services, life sciences, and energy have already seen evidence of value when using Rimay on IBM Quantum hardware, demonstrating the potential of quantum computers to improve workflows as the technology matures.
High Impact Results Across Sectors
Rimay has supported customer projects, including:
Komatsu Peru and NTT Data Latam and Europe used Kipu to get trustworthy predictive maintenance insights from scarce equipment data.
KPMG used Rimay to classify tree species from limited satellite imagery, delivering clearer quantum-enhanced environmental intelligence.
Moeve used Kipu’s technology to analyse thermal imaging and quantum-enhanced leak detection in oil & gas pipelines.
Also Read: Trifork launches Danish sovereign data and AI solution as alternative to US hyperscalers
Organisations have also seen results across a wide range of other industry use cases:
- Credit risk assessment: +5% accuracy vs classical models
- Oil pipeline leak detection: +13% balanced accuracy
- Molecule toxicity prediction: +5–10% accuracy
- Semiconductor fault detection: +20% accuracy
- Drug-induced autoimmune reaction prediction: +7% accuracy
- Company bankruptcy prediction: +4% predictive performance improvement
How It Works: Quantum for Machine Learning
Rimay Quantum Feature Extraction operates as a closed-loop ecosystem where Classical AI and Quantum Computing continuously amplify one another. By mapping complex datasets into a quantum state space, Rimay exposes hidden patterns and high-order correlations that are mathematically invisible to classical computers. At its core, the feature extraction protocol employs digitized counterdiabatic driving to rapidly evolve the system, bypassing typical noise constraints to leverage k-local many-body spin dynamics. These capture both linear variable-to-variable contributions and higher-order multi-correlations, signal that classical models miss and as a result overfit, and feed them back as superior features to maximize ML performance. Results validate the potential value of integrating IBM’s 156-qubit processors across image, tabular, and time-series data over purely classical methods. Rimay converts theoretical quantum dynamics into practical, immediate industrial quantum usefulness that scales along with hardware roadmaps.
SOURCE: Kipu Quantum