Business Problem

• The Alternate Assets Management firm had 43 Major Holdings, with major Portfolio valuing upto $ 100M.
• Having high-end clients, their USP was building a bespoke investment portfolio for each of their clients, which required in depth knowledge of each asset & portfolio, including the historical returns, real time portfolio comparison etc.
• With hundreds of complex documents & financial reports for each portfolio, it took months for their portfolio manager, especially leaders, to understand the offerings & often failed to utilized the true value of these for their clients.

About Corbin Capital

Corbin Capital Partners is a woman-led investment firm specializing in alternative assets like hedge funds and credit investments. They manage client money through various methods and prioritize client satisfaction. As of April 1, 2024, they manage $9.1 billion in assets.

Solution

GoML’s consulting team sat down with different portfolio managers to understand various inputs & features for each portfolio, as well as the data backing them:
A GPT – 4 Turbo powered agent ingested all the portfolio documents (structured, unstructured, images, video & audio transcriptions)
A RAG based engine extracted all the data from these multiple sources and allowed the portfolio managers to interact with this data realtime, to gain inisights, build clear comparisons, understand the portfolio structure, perform complex mathematical functions and get detailed portfolio reports

User Authentication and Authorization

The system ensures that only authenticated and authorized users can access the document repository, maintaining strict security protocols and compliance with regulatory standards. This controlled access helps protect sensitive information and provides an audit trail for monitoring and review.

Natural Language Processing (NLP)

The chatbot employs advanced NLP techniques, enabling users to interact with it using natural language queries. This feature allows users to type questions or statements in everyday language, making the system highly intuitive and accessible.

Efficient Information Retrieval

Once a user submits a query, the chatbot processes it, retrieving the most relevant information from the indexed and processed documents. This rapid retrieval process significantly reduces employee's time searching for information, thus enhancing productivity.

User Interaction Logging

The system logs all user interactions and feedback, which are then analyzed to refine and improve the chatbot’s performance continually. This ongoing learning process ensures that the chatbot becomes more accurate and efficient over time, adapting to users' evolving needs.

Scalable and Automated Processes

The chatbot's implementation automates many of the manual processes previously required for document searching. Its scalable architecture allows it to handle a growing volume of queries and documents without compromising performance, thereby increasing overall operational efficiency.

Architecture

• React.js UI, secure SSL on Linux VM.
• Ingests data from sources like GitHub processes with Vision Parser.
• Azure AI services (OpenAI Vision, Turbo, Embed, AI Search) for data processing and querying.
• Azure Cosmos DB is used for structured data, and SharePoint integration is used for document management.
• User authentication, logging interactions for auditing.
• Scheduler-managed scalable architecture for efficient task execution.
• Analyzes user interactions and feedback for ongoing refinement.