DISCO provides a cloud-native legal solution that simplifies legal hold, legal request, ediscovery, legal document review and case management for enterprises, law firms, legal services providers, and governments. DISCO’s solution is designed to empower legal professionals with cutting-edge tools that streamline their workflows, enhance data analysis, and provide unparalleled insights.
Leading the charge at DISCO is a team of engineers committed to reshaping how legal professionals navigate the intricacies of the legal landscape. Rick Vestal, Director of Engineering, and Matt Hinze, Principal Software Architect, are crafting solutions designed not only to meet but surpass the expectations of legal professionals in an industry where precision and speed are of utmost importance.
Challenge
Legal professionals need access to a comprehensive and diverse set of data to build compelling cases or to respond effectively to legal inquiries. To investigate facts related to litigation or investigation workflows, they need to identify, collect, and analyze relevant information to support these cases. However, legal datasets are inherently dynamic, with volumes fluctuating based on case sizes and information complexities.
DISCO has developed Cecilia AI, an AI-powered platform designed to empower legal professionals by offering immediate access to case facts, reducing manual tasks, and amplifying their ability to deliver results. A standout feature within Cecilia AI is Cecilia Q&A, a purpose-built AI fact expert facilitating seamless natural language interactions. Users can query and analyze large datasets, in an effort to streamline their workflow and enrich their overall experience.
Cecilia Q&A acts on the information stored in a customer's database to deliver insights for legal professionals working with case-related data. During the initial stages of legal discovery, legal experts grapple with substantial volumes of unstructured data like documents, messages, emails, etc., that may contain vital information for ongoing cases. At this crucial stage, Cecilia Q&A is intended to provide a detailed and context-dependent view of the legal content with the goal of providing a more advanced approach than conventional keyword searches, for a more accurate and comprehensive analysis.
Given the urgency for legal teams to swiftly interpret unfamiliar data, DISCO needed a vector database for Cecilia Q&A that could adeptly navigate the intricate complexities of legal information on an expanding scale. Crucially, this vector database had to possess the dual capability of seamlessly ingesting new data while retaining the capacity to address queries related to older datasets.
Solution
The DISCO team recognized the need for a vector database that would support relevant searches across a massive volume of legal data. While other options were available, DISCO selected Pinecone as their vector database partner to support the development of Cecilia Q&A.
As users engage with Cecilia Q&A by posing natural language questions, Pinecone helps facilitate the creation of embeddings to capture the semantic nuances of these queries. The information obtained undergoes enrichment through DISCO's databases for the purpose of creating a comprehensive knowledge base. Through the use of a large language model, the system is designed to generate coherent and relevant responses to complex legal inquiries. By leveraging Pinecone, DISCO sought to achieve efficient vector searches, enabling real-time adaptability, query performance for rapidly and accurately retrieval of information from vast legal datasets, from new and old data.
To enhance efficiency, DISCO wanted to explore an alternative to pod-based structure as it demanded a proactive approach to resource provisioning, often resulting in over-provisioning to mitigate potential surges. Since legal cases are inherently unpredictable, the team was looking for the flexibility to adapt to varying data loads, only paying for the resources used, and eliminating the need for complex pre-provisioning strategies. The team tested the recently available Pinecone serverless, allowing them to eliminate the need for complex pre-provisioning strategies, aligning with the scaling of infrastructure based on customer data ingests.
"Serverless gives us the ability to rethink what we can do with Pinecone for our customers." - Rick Vestal, Director of Engineering, DISCO
result
The decision to transition to Pinecone serverless was prompted by the challenge of managing unanticipated customer ingests. In the pod architecture, the scaling process resulted in doubling resources with each ingest, making it challenging to predict and provision accurately. If a customer ingested a gigabyte of data, it could lead to the creation of one pod. Subsequent data ingests, doubling the volume, would require provisioning two pods, then four, and so on, creating cost inefficiencies. The unpredictability of customer data ingests made it difficult to pre-provision the exact number of pods needed.
With Pinecone serverless, DISCO expects to:
"The transition to Pinecone serverless was seamless. The APIs and contracts remained unchanged for our core interactions, simplifying baseline compatibility testing." - Rick Vestal, Director of Engineering, DISCO
The strategic partnership with Pinecone has laid the foundation for future plans for scalability and innovation. This ongoing partnership underscores DISCO's commitment to staying at the forefront of legal technology, with plans to use Pinecone as a key enabler for future advancements in their offerings.
Matt shares, “Our collaboration with Pinecone has been incredibly beneficial. We've enjoyed direct access to their product team, receiving valuable insights, well-documented guidance, and useful tips. The expertise of the Pinecone team should play a pivotal role in steering us in the right direction, fostering a strong and productive partnership."
"Pinecone serverless opened up possibilities we hadn't considered before and allows us to invest even more in our long-term product capabilities."- Rick Vestal, Director of Engineering, DISCO
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。