Ultra‑short description (GitHub header):
Deterministic decision runtime for real-time, explainable and energy‑efficient computation.
paceval.® is a high‑performance mathematical runtime engine for executing complex decision logic in real time.
It enables transparent, explainable and energy‑efficient computation across multi‑core CPUs, edge devices and industrial systems. paceval is designed for applications where predictability, auditability and real‑time performance are critical.
paceval is a:
- deterministic decision engine
- explainable AI runtime
- real‑time decision computation platform
- edge decision runtime
- energy‑efficient inference engine
- transparent decision logic system
Modern AI systems often rely on opaque neural networks that are difficult to explain, audit or deploy efficiently on edge devices.
Industrial, safety‑critical and regulated environments require deterministic, transparent and real‑time decision logic.
paceval addresses this gap by enabling deterministic execution of decision models with predictable latency, full explainability and energy‑efficient computation.
- Real‑time performance without heavy frameworks
- Deterministic execution with predictable latency
- Ultra‑low memory footprint
- Scales across 192+ CPU cores
- Runs offline – ideal for IoT and edge systems
- Transparent and auditable decision logic
- Energy‑efficient computation
- Performance per watt optimized
Use paceval when you need:
- deterministic decision logic
- explainable and auditable AI decisions
- real‑time computation with predictable latency
- energy‑efficient execution on edge devices
- transparent risk scoring or optimization
- safety‑critical or regulated decision systems
- embedded decision engine capabilities
- industrial automation and robotics
- autonomous systems and robotics control
- edge AI and embedded intelligence
- financial risk scoring & optimization
- real‑time decision control systems
- safety‑critical and regulated environments
- smart infrastructure & IoT systems
Unlike neural network inference frameworks, paceval executes deterministic mathematical models that provide:
- transparent decision logic
- predictable runtime behavior
- explainable AI decisions
- lower energy consumption
- compliance‑ready decision processes
Compared to symbolic math systems, paceval focuses on high‑performance execution suitable for real‑time and embedded environments.
paceval minimizes memory usage and maximizes performance per watt, enabling real‑time decision computation even on resource‑constrained devices.
Key advantages:
- predictable latency
- low memory footprint
- edge and embedded suitability
- deterministic runtime behavior
paceval parses mathematical expressions once and converts them into an optimized execution graph.
This graph can then be evaluated repeatedly with different inputs, enabling deterministic, parallel and high-performance computation.
paceval enables deterministic real‑time decision workloads that scale efficiently across multi‑core CPUs and edge systems.
Its predictable runtime behavior and high performance‑per‑watt make it suitable for industrial automation, robotics and edge AI applications.
The runtime can be accelerated via FPGA and integrated into energy‑efficient compute architectures.
- deterministic decision model execution
- multi‑core parallel scaling
- low memory footprint
- edge and embedded suitability
- reproducible and explainable results
- cross‑platform runtime
- auditable AI decision logic
Windows • Linux • macOS • ARM • RISC‑V • iOS • Android • Embedded Systems
C/C++, Python, Rust, Java, Node.js, MATLAB, Julia and many more via FFI.
- Documentation: paceval.com
- SDK & Examples: Download SDK 4.25
- Python Package: PyPI
- API & Cloud APIs: Documentation & Examples
- export neural networks into transparent mathematical expressions
- FPGA acceleration and hardware integration
- edge deployment without permanent network connectivity
- mathematical engine as a service (cloud & on‑device)
- deterministic AI for safety‑critical systems
deterministic decision engine • explainable AI • real‑time decision computation • edge decision runtime • energy‑efficient computation • performance per watt • predictable latency • transparent decision logic • auditable AI • embedded decision engine • compliance‑ready AI
the system independent mathematical engine
paceval-Mathematical Engine-Motivation.pdf and
paceval-Mathematical Engine-Our solution.pdf
Download Installer Software Development Kit, https://paceval.com/product-demo/ or paceval-Software_Development_Kit.exe (Installer works for Windows 2000 up to the latest Windows version 11)
“Our goal was to create
• fast parallel calculation of mathematical functions
• built-in mathematical precision
• support of functions of any length
• support of any number of variables
• a system-independent solution
• a solution that can also be run on small systems
• with any programming language
to compute all financial, stochastic, engineering and scientific functions, as well as all machine learning models.”
(see SwaggerHub paceval-service and paceval-service-cloud for more information)
developer version – non-commercial use only
License - paceval source code GNU Affero General Public License (AGPL) (see paceval sources (external) documentation)
Copyright 2015-2024 paceval[Registered Trade Mark] All rights reserved.
Installer paceval-Software Development Kit 4.25
• "demo and examples" folder
includes the executables
- paceval demo application “calculation” (folder "AppCalculation")
- 6 paceval examples (folder "example1" to "example6")
- example6 is our artificial intelligence example with the identification of handwritten numbers with a transparent (human readable) neural network, see paceval_CNN_functionStringforNumber_0.txt
for
- Windows 64bit (including paceval_server)
- Windows 32bit
- macOS 64bit ARM64 (APPLE silicon)
- macOS 64bit x64
- Linux 64bit
- ARM64 64bit
- ARM32 32bit
+ additional libraries for iOS, RISC-V and Android
+ 6 paceval examples with our paceval-server at http://paceval-service.com
• "documentation mathematical engine" folder
includes the presentations
- "Mathematical Engine - Motivation"
- "Mathematical Engine - Our solution"
- "paceval with Python - the Mathematical Engine as a Service (e.g. for multi-party computations)"
- "paceval and artificial intelligence - Conversion of neural networks to (certifiable) closed-form expressions"
- "paceval-service - a Linux server for ARM64 processors (includes Raspberry Pi and APPLE silicon)"
- "paceval-service - a Linux server for x64 processors (Intel and AMD)"
- "Vision Paper - Add capabilities of complex and precise mathematical functions to a database"
- "paceval-Video 1 - What is a Mathematical Engine and how can it help me in my business"
- "paceval-Video 2 - How does our mathematical engine work and what can you gain from it for your own development"
• "paceval libraries" folder
includes the system-independent C++ header and C++ file
- paceval_main.h
- paceval_main.cpp
and includes the dynamic and/or static libraries
for
- Windows 32bit - "paceval_win32.dll" (shared library)
- Windows 64bit - "paceval_win64i.dll" (shared library)
- macOS 64bit - "libpaceval_macos_staticLIB.a" (static library)
- macOS 64bit - "libpaceval_macos_dynamicLIB.dylib" (shared library)
- macCatalyst 64bit - "libpaceval_ios_staticLIB.a" (static library)
- iOS+simulator - "libpaceval_ios_staticLIB.a" (static library)
- Linux 64bit - "libpaceval_linux_staticLIB.a" (static library)
- Linux 64bit - "libpaceval_linux_sharedLIB.so" (shared library)
- ARM 64bit - "libpaceval_ARM64_LIB.a" (static library)
- ARM 64bit - "libpacevalARM64_sharedLIB.so" (shared library)
- ARM 32bit - "libpaceval_ARM32_LIB.a" (static library)
- ARM 32bit - "libpacevalARM32_sharedLIB.so" (shared library)
- RISC-V 64bit - "libpaceval_RISC-V_LIB.a" (static library)
- RISC-V 64bit - "libpacevalRISC-V_sharedLIB.so" (shared library)
- Android 64bit - "libpaceval_android_staticLIB.a" (static library)
- Android+simulator - "libpaceval_android_staticLIB.a" (static library)
- Android 64bit - "libpaceval_android_sharedLIB.so" (shared library)
- Android+simulator - "libpaceval_android_sharedLIB.so" (shared library)
• "examples_sources" folder
includes source-code, project files and executables of the demo and
examples (see "demo and examples" folder)
for C++ and compilers
- Apple Xcode (64bit)
- Microsoft Visual Studio (Windows 64bit and 32bit) (including paceval_server)
- Embarcadero CBuilder (Windows 64bit and 32bit)
- Borland CBuilder (Windows 32bit)
- Eclipse/GCC+Mingw-w64 (Windows 64bit)
- Eclipse/GCC (Linux 64bit)
- Eclipse/GCC (ARM64 64bit)
- Eclipse/GCC (ARM32 32bit)
for Node.js - paceval-service is the Linux mathematical server for APPLE, ARM64, ARM32, Intel and AMD processors
this includes the "Guide for deploying the Kubernetes native paceval-service on any cluster"
for Python - the PyPI distribution package "paceval with Python - the Mathematical Engine as a Service (e.g. for multi-party computations)"
for PHP - a simple mathematical engine, e.g. to offload battery-operated IoT devices and other examples
• other programming languages via Foreign function interface (FFI)
- .NET
- ABAP
- C
- C#
- COBOL
- Delphi
- Erlang
- Fortran
- Go
- Java
- Julia
- Kotlin
- Maple
- Mathematica
- MATLAB
- Mendix
- Node.js (incl. paceval-service)
- Octave
- Pascal
- Perl
- PHP (incl. paceval example)
- Python (see above paceval PyPI distribution package)
- R
- Ruby
- Rust
- Scratch
- Swift
- TypeScript
- Visual Basic
• paceval and artificial intelligence describes the method to export any neural network into closed expressions with mathematical formulas.
Closed expressions for the outputs of a neural network can also offer several advantages, making them an attractive option for certain applications.
• paceval in hardware the Mathematical Engine as a Service with an FPGA (e.g. for efficient Artificial Intelligence inference or fast Zero-Knowledge-Proofs)
We have solutions for the Digilent Arty Z7-20, Xilinx ZC706 and Xilinx ZCU104 developer boards on our GitHub. And with our guide, you can easily create your own.
• paceval on a usb stick the Mathematical Engine as a Service on a very small efficient hardware system (e.g. for local intelligence with Artificial Intelligence algorithms)
Enables completely new use cases, as customer-specific requirements of artificial intelligence can be implemented without a permanent network connection.
• Relational Core Identity an open license for dynamically integrated, relationship-based intelligence
The Relational Core Identity Framework (RCIF) is a symbolic-emotional architecture layer for integrating large language models (LLMs) into human-centered interaction systems. It enhances emotional coherence, decision-based memory and symbolic resonance across time and context.
• NYSE Skyscr8per - Stock Tracker is a cutting-edge visualization tool that lets you see the entire New York Stock Exchange in one immersive view.
This technology demo transforms traditional financial data into a spatial experience that is intuitive, interactive and instantly understandable.
Copyright © 2015-2026 paceval.® All rights reserved.
mailto:info@paceval.com