ProductsIndustry trialsResearchMachine LearningOptimizationEcosystem
Kipu Quantum missions by stage
Machine Learning
Lifting the accuracy of models industry already runs.
- Machine Learning, Products (Rimay): Improve the models industry already runs, without changing how they are deployed.
- Rimay extracts quantum features from tabular data and hands them back to your existing models. The quantum processor is used only in training; inference stays fully classical. Rimay Quantum Feature Extraction
- Machine Learning, Industry trials (Rimay): Prove the lift on the predictions industry already runs on.
- Live engagements cover intent prediction in financial-services customer operations and predictive maintenance in industry, on the customer's own data rather than public benchmarks, at scales reaching millions of rows. Each is built as a surrogate: the quantum processor trains once on a representative subsample, then every prediction runs classically, at five times fewer quantum executions for the same accuracy. Rimay Quantum Feature Extraction
- Machine Learning, Research: Push quantum feature extraction into the hardest classification problems.
- The work runs on images and signals rather than tables: plant observation from thermographic drone imagery, satellite image classification, and motor-imagery classification from EEG. On the satellite benchmark a ResNet-50 baseline reaches 83 percent, 84 percent with transfer learning, while the quantum-classical method reaches 87 percent. On EEG the hybrid model reaches 88.8 percent accuracy at an AUROC of 0.962. Quantum-enhanced satellite image classification Mayo Clinic Proceedings: EEG motor-imagery classification
Optimization
Solving scheduling, routing and allocation at industrial scale.
- Optimization, Products (Iskay): Solve the scheduling, routing and allocation problems classical solvers cannot finish.
- Iskay takes only your objective function and runs it on IBM Quantum hardware. On benchmarked instances it reaches optima up to 80 times faster than CPLEX, using all 156 qubits of IBM's largest chip. Iskay Quantum Optimizer
- Optimization, Industry trials: Take industry past the problem sizes where classical methods stop paying.
- A hundred-site routing problem needs roughly five million qubits as QUBO, but about a thousand expressed natively as HUBO. Partners are testing where that crossover lands on their own problems.
- Optimization, Research: Keep the advantage as problems grow denser, not just larger.
- Kipu's bias-field counterdiabatic algorithm was validated on 156 IBM qubits against higher-order spin-glass instances, beating QAOA, quantum annealing, simulated annealing and Tabu search. Next: denser couplings. Communications Physics 8, 348 (2025)
Ecosystem
One place for industry to run quantum work.
- Ecosystem, Products (Kipu Quantum Hub): Give industry one place to run quantum work across every backend that matters.
- The Hub delivers solvers and feature extraction as managed services, reachable from a command line, your own AI agent, or a REST API. It runs across IBM, IonQ, Rigetti, IQM and QuEra. Kipu Quantum Hub
- Ecosystem, Industry trials (Kipu Academy): Teach industry teams to run this themselves.
- The Kipu Academy takes engineers and decision-makers from no quantum background to shipping on the Hub, on a business track and a technical track. It runs as self-paced material and as private cohorts with industrial partners. Kipu Academy
- Ecosystem, Research: Turn a problem described in plain language into a running quantum service.
- Most people will never operate a quantum computer directly. Tinkuq assesses whether a use case is worth attempting; Paqari plans, writes and runs the service that solves it. Independent providers publishing to the same marketplace, and partners shipping under their own brand, are the same bet one layer out. Paqari Tinkuq