How Hercules.ai Used GenAI to Boost Ropers Majeski Lawyer Productivity by 18.5%
Learn Hercules.ai's Integration of a Multi-Modal Database for AI to Boost Query Speed by 100x & Build Efficient GenAI Workflows for Millions of Legal Documents
Increase in productivityIntroduction
Ropers Majeski, an established midsize law firm, has provided high-quality legal advice to clients worldwide for over 70 years.
Ropers Majeski collaborated with Intel, Activeloop, and Hercules.ai to create a multi-layer AI system powered by high-performing 4th Gen Intel® Xeon® Scalable processors.
The solution significantly improved productivity for previously manual tasks such as documenting, filing, timekeeping, storing, and information retrieval while maintaining data security.
The Challenge
Proper filing of documents and emails, information retrieval, and report creation are important knowledge worker tasks at Ropers Majeski offices across the globe. They are also critical for regulatory compliance. Historically, these tasks were done manually, resulting in inaccuracies and substantial time spent on non-billable work.
Ropers Majeski was looking for AI-based automation of these tasks but was concerned about exposing sensitive data. They also wanted a system that would ensure regulatory compliance, cater to the high transaction volume, be reliable, and provide near-instant responsiveness.
A Multi-Layer Gen AI Solution
Hercules.ai partnered with Intel and Activeloop to provide Ropers Majeski with a powerful, reliable, and secure AI-driven solution. This solution is based on a combination of a unique AI application, state-of-the-art data storage, and advanced hardware optimized for large language models (LLM)
Deep Lake
Activeloop Deep Lake provided secure, enterprise-grade storage with advanced role-based access controls to the Hercules.ai solution. Deep Lake enabled storing multiple modalities of data - both embeddings and metadata for millions of sensitive documents at Ropers Majeski. Its serverless Tensor Query engine expedited queries 100x for semantic search and question answering, all within a simple, microservice-free architecture. Finally, thanks to Deep Lake’s performant data loader, PyTorch was able to natively access the data and distribute it automatically across MPI workers.
- Hercules.ai enhanced Ropers Majeski's level of compliance by providing an AI-driven application for assistive documents or email filing, information and insights retrieval, and co-piloting for content generation. The company’s Hercules AI engine systematically labeled and indexed millions of sensitive files while providing powerful generative AI features. Hercules.ai has utilized its proprietary custom distance metrics and semantic search algorithm to provide high accuracy and recall. These were implemented in PyTorch.
- Activeloop Deep Lake provided secure, enterprise-grade storage with advanced role-based access controls to the Hercules.ai solution. Deep Lake enabled storing multiple modalities of data - both embeddings and metadata for millions of sensitive documents at Ropers Majeski. Its serverless Tensor Query engine expedited queries 100x for semantic search and question answering, all within a simple, microservice-free architecture. Finally, thanks to Deep Lake’s performant data loader, PyTorch was able to natively access the data and distribute it automatically across MPI workers.
- 4th Gen Intel® Xeon® Scalable Processors are delivering significant efficiency gains. Hercules.ai has provided Intel-powered state-of-the-art LLMs that meet Ropers Majeski's stringent information security policies.
Results
”A 100x speedup of Tensor Query execution for semantic search and question answering on legal documents. Deep Lake’s minimalistic architecture provided flexibility and light touch installation for our customers without introducing complexity such as adding a microservice. With Deep Lake’s ultrafast data loader, PyTorch was able to natively access the data and distribute it automatically across MPI workers, allowing for highly parallel embedding search.”
Gevorg Karapetyan
Co-founder and CTO of Hercules.ai
Key Technical Achievements
This deployment has yielded significant improvements in efficiency. Utilizing 4th Gen Intel® Xeon® Scalable CPUs, document ingestion, including the embedding computations into Deep Lake using Intel® AVX-512, became 2.3 times faster. Notably, the increase was achieved without compromising accuracy, which is crucial for legal and financial document search and analysis. Additionally, the query speed of Activeloop's Tensor Query Engine improved by a factor of 1.4 when using 4th Gen Intel® CPUs compared to 3rd Gen Intel® Xeon® processors.
18.5% Increase in Productivity
Compared to Baseline Vector or Lexical Search100x Query Speed Boost
For semantic search & question answeringMillions of Multi-Modal Documents for RAG
Within a simple, microservice-free architecture
Customer Outcome
At the end-customer level, Ropers Majeski has seen an average saving of 75 minutes per user per day. With the help of Hercules.ai, most information retrieval, email processing, document filing, and report generation tasks have been automated. Driven by a 2.3x increase in inferencing speed while using an LLM, users have also benefited from near real-time information and insights retrieval. As a testament to the combined power of cutting-edge technologies adopted and developed by Hercules.ai, the holistic productivity of Knowledge Workers escalated by an impressive 18.5%.
Key Learnings
“We are excited about our collaboration with Hercules.ai and Activeloop under the Intel Disruptor Program, in bringing strong value to the end customer Ropers Majeski. The performance and TCO optimization of the Gen-AI solution on Intel’s Latest generation CPUs leveraging in-silicon AI accelerators like Intel AMX, and other IA based software-hardware assets, go a long way to provide consistent benefits and secure offering to the overall solution”
Arijit Bandyopadhyay
CTO of Enterprise Analytics & AI, Head of Strategy Cloud and Enterprise – CSV Group, Intel Corporation
Data Storage as the Cornerstone of AI Excellence
Activeloop's database for AI provides secure, scalable data storage, managing embeddings and metadata for millions of sensitive documents. Our serverless Tensor Query engine expedited queries 100x, showcasing the importance of a robust data storage and management system in ensuring the successful implementation of AI solutions, especially in a highly regulated environment such as legal tech.
We expect to continue to take advantage of the advanced AI technologies to streamline our operations
Maks Agamir
Director of IT at Ropers Majeski
This kind of strategic partnership is just the beginning of integrating generative AI into the legal profession and other highly regulated industries.
Gevorg Karapetyan
Co-founder and CTO of Hercules.aiAs we navigate deeper into the digital age, integrating advanced AI technologies and state-of-the-art hardware presents a powerful opportunity to revolutionize legacy systems in highly regulated industries burdened by vast volumes of data. However, this digital transformation requires overcoming challenges, notably data privacy, security, and regulatory compliance. As demonstrated by Hercules.ai, Intel, and Activeloop, strategic collaborations exemplify how comprehensive, secure, and compliant solutions can revolutionize these sectors. These partnerships foster enhanced efficiency, improved decision-making, and scalability to industries traditionally burdened by manual processes and outdated systems.
“Technology is not our primary business—law is. Attorneys and legal professionals were happier because when their day runs out, instead of sitting there and profiling every email, every document, they just looked and said okay, it’s done. AI did it all”
Maks Agamir
Director of IT at Ropers MajeskiIntel Disclaimers
Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or component can be absolutely secure.
Your costs and results may vary. For workloads and configurations, visit 4th Gen Xeon® Scalable processors at www.intel.com/processorclaims.
Results may vary.
Intel technologies may require enabled hardware, software or service activation.
Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.
Intel® technologies may require enabled hardware, software, or service activation.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.
How Bayer Radiology Uses Database for AI to Disrupt Healthcare with GenAI
Learn how Bayer Radiology, a division of a pharmaceutical powerhouse, used a secure, efficient, & scalable database for AI to pioneer medical GenAI workflows
Read more
IntelinAir faster AgriTech with Aerial Machine Learning Data Pipelines
Learn how IntelinAir, the leading crop intelligence company, transformed 1500 terabytes of aerial imagery into vital insights for farmers with scalable plug-and-play data pipelines with Activeloop and NVIDIA.
Read more
