Innovating in Patent Search

What does it take to build an AI-powered patent search engine from the ground up? In this episode of RallyCast, we go back to where it all started—exploring the early days of IPRally, the challenges of building a unique search technology, and the vision for the future of AI-driven patent research.

Host John Paul Keeler is joined by IPRally co-founders Juho Kallio (CTO) and Juuso Piskonen (Lead Engineer) as they discuss the origins of IPRally, their journey into the world of patents, and the revolutionary use of knowledge graphs to transform patent search. From the first prototypes to today’s cutting-edge AI-powered search engine, they share insights into the development process, the key technical decisions that shaped IPRally, and how they see the future of search evolving.

One of the fundamental breakthroughs was moving beyond traditional Boolean search methods and leveraging structured graph-based AI models to understand the true meaning behind inventions. Juho and Juuso explain how this approach not only improved search efficiency but also created a foundation for future advancements in AI-powered patent discovery.

"We didn’t just want to build another patent search engine. We needed to think bigger. We knew that if we could structure patent data into knowledge graphs, we could create something revolutionary—something that would fundamentally change how patent search works."

In this RallyCast:

• The origins of IPRally: How a chance meeting at a startup event led to the creation of an AI-powered patent search platform.
  ‍
• From idea to reality: The challenges of building a search engine from scratch, developing graph-based AI models, and tackling data processing at scale.
  ‍
• Breaking the mold: Why traditional search methods weren’t enough, and how knowledge graphs transformed the way patents are searched.
  ‍
• The AI search advantage: How using examiner citations as a training dataset ensures accuracy and reliability while maintaining user privacy.
  ‍
• Looking ahead: What’s next for IPRally’s AI technology? The future of AI-powered patent search and the role of distributed computing in scaling machine learning models.
  ‍

The Future of Patent Search with AI

As IPRally continues to push the boundaries of AI-powered search, Juho and Juuso discuss how their compute infrastructure (IPRay)—built on Google Kubernetes and Ray—is enabling unprecedented levels of efficiency and scalability. From cost-optimized machine learning training across multiple global regions to real-time AI improvements, they reveal how IPRally is staying ahead of the curve.

With the foundation of knowledge graphs proving to be a game-changer, the team is looking beyond patents, exploring possibilities for expanding into technical papers and other structured datasets. The goal? To create the most powerful AI-driven search engine for innovation intelligence.

‍

‍