Finster Reimagines Investment Banking and Research with an AI-Native Platform

Summary

Finster accelerates research and analysis for financial institutions, processing millions of vectors while maintaining enterprise-grade security, accuracy, and speed with the help of Weaviate.

Challenge

Financial institutions face overwhelming amounts of data and constant pressure to complete time-sensitive workflows. Analysts can spend hours manually processing multiple documents from different sources for tasks like earnings analysis. This becomes particularly challenging when more than one company in their portfolio reports simultaneously. Manual approaches leave analysts vulnerable to missing critical information, particularly when they are so time-constrained. When considering an AI platform, these organizations need results to be incredibly accurate, fast, and secure.

Why Weaviate

As a rapidly growing fintech startup handling large-scale document processing, Finster saw that Weaviate was the vector database that could meet their strict requirements:

• Robust pre-filtering, reranking, and hybrid search:Precise document analysis with understanding of finance-specific concepts was critical.
• Enterprise-ready platform: Their customer base required sensitive data handling, enabled with multi-tenancy and VPC deployments.
• Scalability: The platform needed to handle millions of vectors while maintaining performance.
• Flexible query patterns: Weaviate could support a diverse range of user queries and task types.
• Growth partner: A supportive technical team was crucial for a quickly growing company.

“Very early on we were fortunate to speak with Byron Voorbach, Weaviate’s Field CTO. He saved us several weeks of iterating on various retrieval methods and was able to guide us towards a specific solution that worked really well. It helped us deliver on the accuracy and consistency our users expect, and we still use the framework of that architecture now,” notes Kilgarriff.

Ultimately, Finster chose Weaviate for its comprehensive feature set, enterprise readiness, and knowledgeable technical team.

From early iterations to enterprise platform

Finster was first built with Weaviate Serverless. “We used Serverless to maintain speed. When we started, there were three of us in the team, so running our own clusters didn't seem like wise use of time,” said Kilgarriff.

As the business grew, so did their requirements. “We made the decision to move from Serverless to Enterprise for a few reasons. We landed more enterprise customers, so we needed more enhanced features around high availability and compression options, along with a closer support relationship and SLAs as uptime and handling higher queries per second became more critical. At the rate we were scaling, Enterprise was also more cost efficient.”

“Since moving to Enterprise from Weaviate Serverless the level of support, responsiveness, and the amount we've learned from the Weaviate team has been hugely valuable,” he adds.

Solution

Finster’s AI-native platform is built on Weaviate, allowing the team to use a variety of large language models (LLMs). Financial data streams in from FactSet and Morningstar with additional real-time data ingestion from SEC filings and Finster’s custom data ingestion pipeline for thousands of companies. The platform delivers:

• Trust and accuracy: Proprietary granular citations at the sentence and cell level enable users to verify generated responses in seconds.
• Finance-specific search: Combining semantic and keyword search using Weaviate’s built-in hybrid search functionality allows for meaningful, industry-aware interactions.
• Time saving task automations: Accelerating complex research end-to-end for tasks like earnings analysis and company profile creation.
• Enterprise-grade security: Finster uses single and multi-tenant deployments based on customer requirements to ensure data remains isolated and secure.

What's Next?

Finster continues to expand its capabilities and customer base at a rapid rate. Their partnership with Weaviate enables them to scale the number of companies they cover, which will more than 5X their vector count in the coming months. Additionally, Finster is exploring cost optimization options including hot, warm, and cold storage and quantization based on guidance from the Weaviate team.

“Damien, our solutions engineer, has been absolutely fantastic. We always go away thinking we've learned about five or six new things just from a 15-minute call with him,” said Kilgarrif. “When you're a growing team moving fast, having experts in certain fields to guide you and bounce ideas off has been really, really helpful. Discussing cost optimization has given us a lot more visibility into how our costs scale as we ramp up our document processing and use of vector databases.”