Oxide AI’s mission is to amplify individuals by giving them access to enterprise-grade AI technology that solves the critical problem of information overload. Oxide’s solutions deliver precise, context-aware insights at the right time, empowering people to make smarter, faster decisions supported by trusted and transparent AI.
THEIR GOAL
Extract insight from the flood of financial data
The volume, velocity and complexity of financial information far exceed human comprehension, making smart decision-making nearly impossible for the average investor. Oxide wanted to use large language models (LLMs) to overcome this information overload by researching markets and generating valuable insights for investors.
THEIR SOLUTION
Llama-powered finance app cuts through the noise
Oxide developed Oxogen AI, an intelligent app that scours the web for quantitative knowledge like trading data, SEO filings and quarterly reports, plus qualitative information like patent applications, news and social media feeds. The app uses multiple lightweight Llama models to extract critical information and generate valuable financial content.
THEIR APPROACH
Multi-agent engine fuses quantitative and qualitative insight
Qualitative and quantitative information come together in EvoQ™, Oxide’s proprietary AI engine. EvoQ uses multiple reasoning AI agents operating at high speed to research entire markets quickly. To ensure accuracy, Oxide used LoRA to fine-tune lightweight Llama models for data extraction, classification and content generation.
THEIR SUCCESS
Open-source Llama put Oxide AI in control of their data and performance
Shifting to open-source Llama helped address key challenges around privacy, domain-specific fine-tuning and production efficiency, all while achieving major cost savings. Small Llama models matched the accuracy of GPT but used just a fraction of the computing resources.With Llama powering its generative AI components, Oxogen AI is positioned to take personal financial services far beyond traditional investing solutions and deliver on a global scale.
“Llama stood out as an open, transparent model that we could fine-tune using our proprietary datasets. The smaller Llama model was also a great fit for our production needs, allowing us to maintain quality while optimizing compute cost and speed. Beyond performance, the fact that Llama is backed by Meta gave us confidence in long-term support and innovation.”
Kateryna Wikström, CPO and Designer, Oxide AI
“Our hybrid-AI approach combines highly refined domain-specific data, computational AI and Llama LLM support into a powerful, efficient system. It enables us to deliver precise, high-quality, context-aware outputs at a fraction of the cost, achieving results that would otherwise take hundreds of hours of manual research.”
Lars Hard Chief AI Officer, Oxide AI
“Meta is constantly updating Llama models and making them available across major cloud platforms quickly, so integration into our workflows is smooth and scalable. Regular improvements, easy integration and flexible fine-tuning made Llama the natural choice for building a trustworthy, efficient and domain-specific generative AI layer in our platform.”
Shopify | Llama case studiesShopify uses Llama to generate product pages, localize content, and automate support, helping developers scale workflows and save time.Read more
ConsumerTech
Scribd, INC | Llama case studiesDelivering faster, cheaper and more accurate results with a Llama-powered AI content discovery assistant.Read more
Tech
Exati | Llama case studiesTransforming support for smart city platform clientsRead more
Llama-powered Oxogen AI delivers quantitative and qualitative financial data to investors.
Llama powers custom research agents in the Oxogen AI app.
With LoRA, smaller models can produce highly accurate, domain-specific results that are as good as or better than the results produced by general-purpose models hundreds of times larger. Using smaller models conserves computing power, which allows Oxogen AI to scale up without incurring massive computing costs.
Lightweight Llama models with LoRA power multiple research agents and content generation.
40% reduction in production costs for LLM-generated content and extractions
37% faster average sequential response times than OpenAI GPT-4 while maintaining comparable quality
95% accurate results after a month of LoRA tuning
*All results are self-reported and not identifiably repeatable. Generally expected individual results will differ.
Stay up-to-date
Our latest updates delivered to your inbox
Subscribe to our newsletter to keep up with the latest AI updates, releases and more.
