Highlights
- Custom Web Development and Extensive AI/ML Expertise: developing a powerful web platform from the ground up, implementing sophisticated features including advanced document search functionality and automated document processing of 30,000 documents.
- Increasing Client’s Customer Base by 40%: delivering an innovative software solution that resulted in 20+ enterprise customers using the services of our client regularly.
Client
The client is Legal Shelf, a Mexico-city-based B2BLegaltech provider, established by legal professionals and catering to the Latin-American market. The customers of our client include some of the Fortune 50 Transnational Companies operating in Mexico in the Telecommunications, Tech Manufacturing, and Tech Consulting sectors, as well as a growing roster of mid-market Mexican companies in the Manufacturing and Retail industries.
Product
The product is a B2B LegalTech solution, a complex web-based software platform that allows customers to comprehensively and securely manage and exchange legal documents. The platform enables collaboration on legal documents in real-time, using strong encryption for all document-related operations. The solution also makes it possible to share documents with external participants by inviting them as guests temporarily.
There is also a critical AI-powered part of the product, aimed at automating document management processes with the Named Entity Recognition (NER) technology. It recognizes and extracts from document text such entities as participants, expiration dates, and contract durations. This functionality also enables entity relations extraction, subtype identification of documents, and the identification of renewal options within the documents.
Overall, the product comprises the following functional areas:
- A digital library with simple and advanced search
- Automated document management
- Simple and advanced document-signing process
- Alerts and notifications
- Diverse tasks management
- Roles management with a complex hierarchy.
Goals and objectives
- Develop a Cutting-Edge Feature-Rich Web Application: Creating a custom web application for managing legal documents tailored to businesses across different industries, allowing them to securely store, exchange, and sign documents.
- Integrate the Product with External Systems: Delivering a holistic, full-service system by integrating the product with three external applications, and communicating with 3rd-party vendors directly since not having any APIs available for the integration.
- Gradually Expand the Product: Ensuring that the product evolves with the client’s business needs, supporting scalability and new features over time. During the development process, changes included adding push notifications, more user roles, and an admin panel.
- Implement Advanced AI/ML Functionality: Establishing a continuous environment for the data labeling process, implementing an information retrieval ML model to expedite document labeling and parsing, integrating Optical Character Recognition (OCR) for automated document reading, and applying spell correction to enhance the accuracy of OCR-recognized files.
Project challenge
- 5-Month Deadline: meet the set timeline for the initial version of the product, enabling the client to kick off the marketing campaign on time.
- Advanced Search as the Flagman Feature of the Product: implement a corresponding number of data fields for the different document types and subtypes to deal with numerous diverse criteria for users to search. This feature was prioritized by the Product Owner as a key market advantage.
- Three Language Support: implement full language support for Spanish, Portuguese, and English to maximize the coverage of the product for the LATAM market. We were also discussing and researching the opportunities to add another third-party service for more languages, if needed.
- Accurate Extraction of Key Information from Legal Documents: leverage our proficiency with AI/ML to automate the extraction of critical information from legal documents with high precision, including information on participants, expiration dates, contract durations, entity relations, document subtypes, and renewal options.
- Integrated Data Labeling Environment: ensure that the data labeling process for the AI/ML solution and the usual document processing take place simultaneously without any disruptions to the critical functionality.
Solution
To implement the project, SPD Technology assembled a dedicated project team consisting of:
- Project Manager
- Software Architect
- Business Analyst
- UI Designer
- Two Backend Developers
- Two Frontend Developers
- 2 AI/ML Engineers.
Legal Shelf already had a legacy system written in PHP that did not satisfy the business demands anymore. Our team chose to build a new, serverless solution using AWS Lambda due to its high availability, out-of-the-box scalability, and versioning support facilitating easy deployment.
Our database choice, PostgreSQL, was driven by its suitability for both OLTP and OLAP functionality. AWS S3 was selected as the storage solution for documents, contracts, and eSignature provider metadata, offering secure and durable storage with rapid access.
Migrating data from the client’s legacy storage presented a significant challenge due to the intricacies of mapping legacy data models and the scale of the client’s data. Intensifying communication and implementing an automatic script proved highly effective, streamlining the migration process and eliminating associated challenges.
The implementation of a fast and intuitive search of the document library was probably our biggest technical highlight of the entire project. We decided not to use external solutions but rather developed the full-text search engine with PostgreSQL in-built. After we made some customizations, this choice showed great results, both in quality and performance. Our search function could find fuzzy matches and provide accurate results ranking.
After the core functionality was completed against a strict deadline, our team shifted focus to enhance the product’s features with Artificial Intelligence and Machine Learning. The ultimate goal was to automate manual document processing, making it easier for users to manage and organize documents. This involved developing features to categorize documents, extract key information, and suggest actions based on the document’s content.
We leveraged Natural Language Processing (NLP) techniques like Named Entity Recognition (NER), entities relation extraction, information retrieval to automatically identify and extract information from legal documents, including participants, expiration dates, and contract durations. In addition to extracting individual entities, we added functionality for the system to determine relationships between entities. This involved identifying which participants are associated with which contracts, as well as identifying renewal options based on contract content. Our system also classifies documents into different subtypes based on their content, including identifying whether a document is an NDA, a contract, or others.
We used the documents processed by the markers as part of their current work as the main data source for the AI/ML solution, integrating this part of the system into the existing client infrastructure via API. Our team ensured the quality and reliability of the data with regular data sample reviews. To validate the functionality and accuracy of our custom AI/ML solution, our team conducted a combination of regression, accuracy, robustness, UAT, and performance testing.
-
AWS Lambda -
AWS RDS -
AWS S3 -
AWS CloudFront -
GitHub Actions -
Terraform -
Auth0
-
Java -
Quarkus -
GraalVM -
Hibernate -
PostgreSQL
-
React -
TypeScript
-
OpenAI
-
Shortcut -
GitHub -
Slack -
Figma
Our results
We successfully implemented the new mission-critical software solution, allowing Legal Shelf to significantly strengthen and dramatically expand its business presence in the national market.
- Modern AI-Powered Web Platform: A highly customized software solution we created resulted in significant cost savings, groundbreaking functionality, and limitless scalability compared to an existing legacy platform.
- Enhanced Customer Experience and Stronger Brand Image: The new platform allowed our client to make the services considerably more accessible for the customers. The client was able to lower prices while offering wider functionality and better performance up to the latest industry standards, and ultimately became the leading digital solution for document management in Latin America.
- Improved Speed of Document Processing: Our AI/ML model helps the employees of a client to extract needed information from documents much faster. Human experts only need to correct and check the model’s results. The AI/ML models are in the constant enhancement process with continuous data labeling, so the speed of document processing will only improve over time.
Highlights Client HaulHub is a B2B2C enterprise that offers digital platforms to the transportation construction sector. The company focuses...
Explore Case