Contract Analysis with an AI-Driven Chatbot

Traditionally, the institution relied on unstructured PDF formats to store its library of contracts. This method introduced many inefficiencies, including the time it took to retrieve data and the risk of errors. Because these contracts often contain highly confidential and proprietary information, data security was also a critical concern.

Six Feet Up engineered an advanced AI-driven chatbot designed to automate the extraction of data from PDF contracts using straightforward queries. This innovation streamlined the client’s operational processes, speeding up data access, enhancing accuracy, and minimizing errors.

The chatbot was stored on a private server, ensuring the company’s data would remain confidential and secure, adhering to the highest standards in data security required in financial operations.

The chatbot utilized a curated suite of technologies:

• Establish a Secure and Scalable Backend: The project began with Python and AWS — foundations that provided a stable and scalable infrastructure for backend development and cloud hosting. This setup ensured a secure and reliable framework that was essential for the sophisticated tools and processes implemented later.
• Retrieve Data Efficiently with Vector Databases: To manage and retrieve large amounts of contract data efficiently, Vector Databases with Weaviate and FAISS were integrated. These databases enhanced the system's ability to index and retrieve data quickly, significantly reducing the time needed to respond to queries and streamlining the data interaction process.
• Process Data Securely with LLMs: The integration of local Large Language Models (LLMs), specifically Zephyr models, was critical. These models were the interpreters of complex natural language queries, discerning the nuances of legal language in proprietary contracts. By processing data locally, these models ensured that the sensitive information remained secure, providing answers with both high relevance and precision, tailored to the unique needs of the client.
• Enhance Flexibility with OpenAI Models: OpenAI Models were introduced as an optional layer to provide additional flexibility and cognitive capabilities. This implementation allowed local LLMs to tap into OpenAI’s powerful capabilities as needed, offering scalability and enhancing the system’s ability to adapt to future requirements.
• Develop a User-Friendly Interface: The project also focused on usability through frontend integration via Chainlit. This approach resulted in an intuitive and user-friendly interface that made the sophisticated system accessible to all users, regardless of their technical background.
• Customize Data Utilization for Context-Aware Responses: Six Feet Up developed custom algorithms to leverage client-specific data and ensure that the AI system’s responses were both accurate and highly relevant to the client’s specific context. Data utilization improves the quality of responses which leads to better decision-making.

Several features set this chatbot apart from standard solutions:

• Custom AI Implementation: The use of local LLMs on AWS highlighted our commitment to data privacy and tailored solutions that respect client-specific requirements.
• Advanced Data Handling: The integration of vector databases like Weaviate and FAISS allowed the client to manage and retrieve large data sets efficiently.
• Integration Flexibility: The flexible integration of OpenAI models into the system offers future-proofing benefits, allowing the client to tap into emerging AI advancements seamlessly.