Supercomputing without Supercomputers

Enabling new frontiers in science and technology by designing new algorithmic foundations

Supported by:

In collaboration with:

Expanding the boundaries of computation

Cambridge Brain is a deep technology company that designs next-generation computing architectures to drastically improve computational efficiency. Unlike many computer architecture initiatives focused primarily on compatibility with current computational models, our emphasis is on creating intellectual property that supports future computing technology stacks. In particular, our solutions address critical bottlenecks affecting high-value computing applications, which currently face challenges in scalability due to constraints in compute power, data handling, or associated costs. The range of potential applications expands alongside technological development and presently includes industrial constrained optimization software (TRL 3-4), AI boosters such as synthetic data generators, and in the future, we aim to extend our capabilities toward data-efficient machine learning.

Our focus on designing performance improvements by one or more orders of magnitude against dominant existing methods.  The objective of our technology is two-fold:

  • Compute Efficiency / Supercomputing: to significantly reduce the number of computations required for critical applications (in optimization and learning), enabling supercomputing without supercomputers.

  • Scale / Breaking Bottlenecks of Progress: to enable new frontiers in science and technology which are held back due to computational cost or complexity of solving larger scale problems.

We have conducted feasibility studies across multiple applications including DNA-assembly, route-optimization, scheduling, and constrained combinatorial optimisation. For more information, please contact us. 

Special call for open-discussions

We are seeking feedback from molecular engineering experts on our proposed feasibility study. This study focuses on the application of our computing engines to identify novel drug candidates in scenarios where data is too sparse to train effective ML models. This approach will compliment current AI models by addressing data-sparse structures which are currently not solvable. The goal of the feasibility study is to determine whether our solver's new abilities, which enable solving problems of orders of magnitude larger than previously possible, can unlock new capabilities in drug candidate discovery.