IPRally: Vespa Architecture Audit for a Production Patent Search Platform
As search systems mature, the hardest questions are no longer just about relevance. They are about whether query execution, infrastructure, and data modeling are making the right trade-offs for performance and scale.
Searchplex worked with IPRally to audit a production patent search platform built on Vespa, reviewing architectural decisions across search behavior, data representation, infrastructure, and performance optimization.
Client
IPRally builds what it describes as “the most accurate patent search engine on the planet.” Their SaaS platform serves patent professionals, IP researchers, and competitive intelligence teams who depend on precise retrieval across large global patent datasets.
The platform is built on Vespa and operates as a production system serving real client workloads.
The situation
IPRally had already built a sophisticated patent search platform on Vespa and the system was working well in production. The engineering team had made deliberate architectural decisions and the platform was successfully serving customers.
As the company prepared for the next phase of growth, the CTO wanted an independent expert review before scaling further.
The goal of the engagement was not to fix a broken system, but to answer a different set of questions:
- Are we using Vespa in the most effective way?
- Are there implementation patterns that could become expensive or difficult to maintain at higher scale?
- Are there capabilities in Vespa that we are not yet using that could improve performance or simplify the system?
Searchplex was engaged to conduct a structured architecture audit of the Vespa deployment and provide recommendations across several areas of the search stack.
The review focused on key parts of the system including:
- custom search behavior implemented in the Vespa application
- wildcard search functionality
- patent data representation and storage
- infrastructure and deployment choices
- debugging and performance analysis practices
This pattern is common in mature search systems. Once a product is working successfully in production, the next level of improvement often comes from validating architectural trade-offs and identifying the highest-impact optimization opportunities.
What we reviewed
Searchplex conducted a structured audit of IPRally’s Vespa-based patent search system hosted on Google Cloud Platform.
The review examined how the platform used Vespa across several core areas.
Custom search logic
We evaluated how custom search logic was implemented within the Vespa application and where the current approach aligned well with Vespa best practices versus where complexity could potentially be reduced.
Wildcard search behavior
Wildcard queries can be computationally expensive and difficult to optimize cleanly. The review examined how these queries were handled and identified opportunities to improve their performance characteristics.
Patent data representation
Patent search systems have unique data modeling requirements. We reviewed how patent data was represented and indexed within Vespa to identify opportunities for improving efficiency and reducing processing overhead.
Infrastructure and deployment
The system infrastructure was reviewed to assess trade-offs between different deployment approaches, with the goal of supporting long-term scalability and operational efficiency.
Debugging and performance analysis
We assessed the existing methods used to investigate search behavior and performance issues and provided recommendations for improving observability and troubleshooting.
The engagement produced both immediate suggestions and longer-term architectural guidance aimed at improving performance, scalability, and maintainability.
Architecture takeaway
This engagement reflects a pattern frequently seen in mature search platforms:
- Production search systems can still contain hidden inefficiencies in query execution, data modeling, or infrastructure configuration.
- Architecture audits are most valuable after initial product success, when teams want to validate trade-offs rather than simply ship new features.
- Patent search systems place pressure on both retrieval quality and system design, making efficiency, observability, and architectural clarity especially important.
- Independent expert reviews help teams confirm which design decisions are sound and where targeted improvements will have the greatest long-term impact.
Outcome
The audit delivered a comprehensive written review of the Vespa implementation along with a prioritized roadmap for future optimization.
Some of the expected benefits identified through the review included:
- reduced wildcard query latency
- lower infrastructure resource utilization
- elimination of inefficient post-filtering processes
At the time of the review, these recommendations had not yet been implemented. The primary outcome was architectural validation — confirming that the core system design was sound while clarifying the trade-offs involved in future improvements.
The optimization roadmap provides IPRally with a structured path to implement performance and efficiency improvements as the system continues to scale.
Searchplex delivered a thorough and well-structured analysis within the agreed timeline. The report is comprehensive and addresses all the areas defined in the scope of work. Ravindra communicated well with us, was receptive to our feedback, and provided good suggestions throughout the work.
Patent Search / Intellectual Property Technology
Production patent search system built on Vespa
Vespa Review & Optimization (Search Stack Audit)
Custom search behavior, wildcard search, patent data representation, infrastructure, debugging and performance
Architectural validation and prioritized optimization roadmap
Query latency, infrastructure efficiency, streamlined search processing
Searchplex conducted an expert audit of IPRally’s Vespa-based search platform to validate architectural decisions, uncover inefficiencies, and deliver a prioritized roadmap for improving performance, scalability, and operational efficiency.