UnQi

Alchemy of Uncertainty

Research and development of open-source software for decision-making under uncertainty and scientific computing

Our Path

At UnQi, we focus on developing robust, open-source software for probabilistic machine learning and scientific computing, specifically targeting complex decision-making under uncertainty. Our path is fundamentally research-driven and hype-free: we build foundational tools, expose them to real-world industrial challenges, and openly publish the insights we gather. This transparent, experiment-led approach not only enriches the scientific community but also forms the bedrock of our business model, allowing us to provide deeply informed, customized extensions of our software to meet the unique and specific needs of our clients.

Our Realm

Uncertainty-Aware Decision Systems

Advanced decision-support systems that quantify uncertainty, evaluate competing options, and recommend robust strategies using Bayesian inference, stochastic optimization, and scenario analysis.

Probabilistic Information Synthesis

Intelligent data fusion platforms that combine signals from multiple sensors, agents, or information streams to produce coherent situational awareness with quantified uncertainty and reliability estimates.

Probabilistic System Modeling

Using probabilistic programming to create probabilistic models of industrial systems, logistics networks, or operational environments that simulate real‑world uncertainty, enabling decision‑makers to test strategies and optimize outcomes before deployment.

Hybrid Intelligence Decision Platforms

Decision architectures that combine probabilistic modeling, domain experts, and LLMs to enhance strategic and operational decision making.

Continuing a Tradition of Open Learning

We are the result of communities, mentors, and institutions that shared advanced knowledge with us at no cost. Following their example, we are committed to organizing courses and workshops in probabilistic machine learning and scientific computing for universities and institutes, as long as they remain fully free for everyone who attends.

Projects / Publications

Zagros

A Parallel and Distributed Optimization Framework for HPC Clusters

Zagros is a black-box optimization framework designed for HPC clusters. It lets users express their own search algorithm in terms of a language called Dena. After designing the search algorithm you can execute it on HPC clusters while leveraging hybrid parallelism. Dena provides various components for designing custom optimizers in an efficient way.

Express your arbitrary complex search algorithm in Dena
Execute the designed search algorithm on multiple nodes
Enjoy hybrid parallelism: Multi-Threading (TBB) + Message Passing (MPI)
Block optimization for memory-intensive optimization methods

HPC Optimization Parallel Computing Distributed Computing Scientific Computing

C++ MPI TBB

GitHub Homepage

@software{RockyML,
    author = {Asadi, Amirabbas},
    doi = {10.5281/zenodo.7612838},
    month = {2},
    title = {{RockyML, A Scientific Computing Framework for Non-smooth Machine Learning Problems}},
    url = {https://github.com/amirabbasasadi/RockyML},
    year = {2023}
}