Weexplore new ways for machines to learn, reason and create.
Oxide AI spent its initial years developing the core AI/ML capabilities that would eventually power its proprietary Polychaos® platform. Rather than pursuing research as an end in itself, our focus has always been on creating foundational AI capabilities that can be translated into scalable products and real-world impact.
Our research focuses on three areas:
Innovative approaches beyond SGD and conventional deep learning topologies.
Operating outside the bounds of training data.
Refinement of high-capacity and high-accuracy domain-specific language models (LLM/SLM).
Research doesn't happen in isolation.
Advancing financial AI with open-source large language models. We explore advanced language capabilities to tackle complex user queries and enhance agentic solutions.
Utilizing SOTA open models to research and develop advanced AI capabilities addressing complex challenges in financial sector.
Adapting models to finance-specific tasks, including query decomposition, text generation and entity extraction.
Using the computing infrastructure at Arrow's labs to run experiments, evaluate models and scale research.
Oxide AI’s research in automated program synthesis is closely tied to the Faculty of Science at Lund University, as part of an advanced program in Evolutionary Ecology.
Bio-inspired AI approaches to optimization and adaptation.
Connecting academic research with industry challenges and real-world applications.
Advanced students can dive deeper in AI and ML through their Master's Thesis projects.