Fission Labs’ FloTorch Powers RAG Framework-Based AI Agent for BrightReturn
Customer Introduction
BrightReturn is a comprehensive platform serving CPAs and law firms, industries known for their heavy reliance on up-to-date regulatory, tax, and legal information. The firms operate in highly dynamic environments, where staying compliant with evolving laws and regulations is critical to their success.
Challenge Overview
CPAs and law firms face significant challenges in keeping pace with rapidly changing tax rules, government policies, and legal precedents. This information is scattered across numerous government websites, subscription-based services, and private resources.Researching this information not only consumes on an average two hours per day but also distracts professionals from their core responsibilities, incurring an estimated cost of $5,000 daily per firm.
For law firms, the complexity was further compounded by the need to:
- Monitor legal precedents across multiple jurisdictions.
- Stay aligned with ethical standards and compliance requirements.
- Quickly retrieve and analyze legal clauses in light of new updates.
Without addressing these challenges, firms risked non-compliance, financial penalties, and reduced operational efficiency.
Solution Overview
BrightReturn partnered with Fission Labs and its FloTorch platform to address the intricate challenges faced by CPAs and law firms. The FloTorch platform's expertise in AI experimentation and productionization, powered by AWS, laid the foundation for deploying a RAG framework-based AI agent, delivering a scalable and efficient solution.
- Knowledge Base Construction
- Aggregated data from 23 distinct sources, including federal and state-specific regulations, IRS publications, and ethical standards for CPAs and law firms.
- Customized for all 50 U.S. states, addressing state-specific tax laws and regulations.
- Consolidated multiple data formats (e.g., PDFs, APIs, HTML) into a unified structure to enhance query efficiency and user experience.
- Designed modular workflows for incremental updates, enabling seamless integration of new resources or data corrections.
- Integrated metadata tagging for improved search ability and contextual relevance across diverse datasets.
- Leveraged FloTorch’s Knowledge Base Management tools to enable efficient chunking, indexing, and embedding of data.
- Built a robust and queryable database using Amazon S3 for storage and Amazon Elasticsearch Service for indexing and retrieval.
- Real-Time Retrieval and Updates
- Deployed a monitoring system using AWS Lambda, Amazon CloudWatch, and FloTorch’s Vector Database Gateway to detect updates in laws, policies, and rules from government and private sources.
- Incorporated a feedback loop for users to report discrepancies, ensuring continuous improvement and trust in the knowledge base.
- Ensured manual validation for every detected update to maintain accuracy and reliability.
- Generative Insights
- Used Amazon SageMaker and FloTorch’s Inference Gateway to integrate advanced language models capable of contextualizing retrieved data and generating precise, actionable insights for users.
- Scalable Infrastructure
- Leveraged Amazon EC2, Amazon RDS, and FloTorch’s optimization modules for processing and managing large-scale data, including over 57,000 PDFs and 62,345 legal cases.
- Enabled weekly updates dynamically with future-proof automation.
- Applied FloTorch’s cost optimization features, such as batching and caching, to enhance system efficiency.
- Fission Labs Expertise
- Fission Labs provided strategic guidance and technical expertise during the pre- and post-implementation phases. Their focus on optimizing workflows ensured seamless integration of FloTorch’s tools with BrightReturn’s existing platform.
- AWS Service Integration
- FloTorch, built with AWS-native solutions, ensured high availability and scalability by leveraging Amazon SageMaker, AWS Lambda, and Amazon Elasticsearch Service. These services provided the backbone for dynamic knowledge retrieval, data indexing, and advanced AI capabilities.
"At BrightReturn, we leverage AI to transform workflows for CPAs and law firms. Implementing FloTorch’s capabilities has enhanced AI experimentation, indexing, and retrieval. FloTorch’s focus on open-source experimentation and scalable production enables us to deliver efficient, compliant AI solutions. Its advanced features, including the Inference Gateway and Vector Database Gateway, optimize our RAG framework-based AI agent, while AWS cloud integration continues to redefine AI solutions for the professional services industry." - Gaurav, Chief Product Officer, BrightReturn
Benefits Delivered to Customer
The implementation of BrightReturn’s AI-driven solution yielded measurable benefits:
- Efficiency Gains
- Reduced research time by 75%, saving firms over $3,750 daily.
- Enhanced Accuracy and Compliance
- Provided a single source of truth for laws, policies, and legal updates, significantly reducing errors.
- Improved Customer Satisfaction
- 90% of users adopted the feature within three months of its launch.
- Feedback highlights:
- "The ability to query a centralized knowledge base saves us hours every week." — Partner at a leading CPA firm.
- "Staying updated with legal precedents across jurisdictions has never been this seamless." — Senior Attorney at a top law firm.
- Scalable Impact
- Indexed tax laws for 38 states and legal precedents dynamically.
- Integrated real-time updates for IRS newsletters and legal digests.
About Fission Labs' FloTorch Platform
FloTorch, created by Fission Labs, is designed to expedite the proof of concept journey for Generative AI workloads. It enables customers to optimize RAG implementations through systematic experimentation and validation, ensuring secure, automated, and cost efficient deployment within their AWS infrastructure. Key features include hyperparameter optimization, vector database enhancement, and LLM integration.
The platform is open-source, powered by AWS, and available through the AWS Marketplace.