Matlogica provides a toolkit that utilizes Automatic Adjoint Differentiation (AAD) and Just-In-Time (JIT) compilation to enhance the performance of simulations and repetitive calculations on existing CPU architectures, achieving speed improvements of 35x to 150x without extensive code rewrites. This technology enables significant reductions in cloud computing costs and accelerates model development, allowing clients to optimize their analytics processes efficiently.
Funding
Funding not disclosed
Founders
Product
Problem
Complex model development and repetitive calculations often suffer from performance bottlenecks on existing CPU architectures, leading to increased cloud computing costs and prolonged model development cycles. Traditional methods for optimizing code require extensive rewrites and can be time-consuming and complex.
Solution
Matlogica offers a toolkit that leverages Automatic Adjoint Differentiation (AAD) and Just-In-Time (JIT) compilation to significantly enhance the performance of simulations and repetitive calculations. The toolkit automatically converts object-oriented code into bare-metal machine performance, designed for cross-platform execution. This approach enables substantial reductions in cloud computing costs, accelerates model development, and improves hardware efficiency without requiring extensive code refactoring. Matlogica's solution allows developers to focus on model creation while the toolkit handles performance optimizations and sensitivity computations. The toolkit supports mixed-mode execution across C++ and Python, allowing for flexibility in defining custom payoffs.
Target Audience
The primary target audience includes quantitative analysts, financial engineers, and data scientists in financial institutions and other industries who develop and deploy complex models requiring high-performance computing.
Features
- Code Generation AAD™: Converts object-oriented code into optimized machine code on-the-fly.
- JIT Compiler: Custom-built compiler enables AAD sensitivities to be computed faster than the original calculation.
- CPU Vectorization: Unlocks native CPU vectorization (AVX2 and AVX512) for improved performance.
- Multi-Threading: Enables safe multi-threading for parallel processing.
- Cross-Platform Execution: Supports execution across various hardware platforms.
- Secure Binary Kernels: Facilitates secure and scalable cloud compute.
- Automated Integration and Debugging Toolkits: Simplifies integration and debugging processes.