Get in touch
Get in touch

AI-Powered Investment Intelligence Assistant for Portfolio Managers

A leading wealth management firm sought to enhance investment research and portfolio decision-making by leveraging AI-driven financial data retrieval and risk analytics. Portfolio managers and financial advisors needed fast, accurate access to investment fund reports, risk assessments, and performance comparisons, but traditional methods were manual, slow, and fragmented.
To address this, ekSource developed a secure, AI-powered Investment Intelligence Assistant, enabling portfolio managers to query structured and unstructured financial documents and receive precise, document-backed responses in real-time. The solution integrates Pinecone as the vector database, BGE-M3 as the self-hosted embeddings model, Llama 3.1 405B for secure LLM inference, and an AWS-based AI pipeline for real-time document ingestion. The web-based chatbot, built with ReactJS, provides an intuitive and interactive experience for financial advisors.
By leveraging ekSource’s AI expertise, proprietary document processing frameworks, and deep experience in financial AI solutions, the PoC was delivered 40% faster and at a significantly lower cost compared to competing AI providers.

Business Challenge

Portfolio managers and investment analysts rely on multiple data sources—including fund prospectuses, earnings reports, SEC filings, and historical performance data—to make informed investment decisions. The existing process of retrieving and analyzing this data was inefficient due to:
  • Manual Data Extraction & Analysis – Advisors spent hours searching through documents to extract relevant financial insights.
  • Fragmented Information Access – No centralized AI-powered knowledge base for retrieving investment intelligence across structured and unstructured data.
  • Lack of Contextual Search & AI Reasoning – Traditional search tools used basic keyword matching, which lacked semantic understanding and contextual relevance.
  • Strict Data Privacy & Compliance Requirements – The firm required an on-premises AI model, ensuring no data leaves its AWS environment.
The client needed an AI-powered solution that could:
  • Retrieve and summarize fund performance and risk analysis from multiple reports.
  • Maintain contextual awareness across multi-turn investment-related queries.
  • Provide document-backed responses to minimize AI hallucinations.
  • Run entirely within their AWS infrastructure for full data security and compliance.

Solution

ekSource designed and implemented a modular AI-powered Investment Intelligence Assistant, combining secure LLM inference, advanced vector search, and real-time document processing.
AI Model Deployment & Secure Inference
  • Llama 3.1 405B was deployed in a closed AWS environment, ensuring zero external API dependencies.
  • The model ran on AWS Inferentia2 (Inf2) instances, optimized for low-cost, high-performance inference.
  • The entire pipeline remained within AWS, maintaining strict compliance with financial data regulations.
Secure Investment Document Processing & Vectorization
  • BGE-M3 was used as the embeddings model, ensuring efficient vector representation of investment documents.
  • Pinecone stored the vectorized data, enabling fast, semantic search retrieval of fund prospectuses, financial reports, and SEC filings.
  • The AI pipeline supported metadata-based filtering, allowing users to retrieve insights based on stock ticker, date, and industry.
Real-Time RAG Pipeline for Investment Research
  • AWS Lambda + S3 + EventBridge handled real-time document ingestion, automatically processing new fund performance reports.
  • A retrieval-augmented generation (RAG) pipeline ensured that all AI responses were strictly document-grounded, reducing hallucinations.
  • LangChain powered the AI query orchestration, enabling multi-turn financial Q&A and context-aware investment comparisons.
Interactive Web-Based Investment Chatbot
  • A ReactJS-powered chatbot UI allowed portfolio managers to query investment data, retrieve documents, and analyze fund risk metrics.
  • AWS API Gateway + Lambda backend enabled real-time AI response generation.
  • The system provided interactive fund comparison tools, allowing advisors to evaluate risk-adjusted returns and fund performance metrics instantly.
AWS-Optimized Infrastructure for Cost & Performance
  • AWS Inferentia2 (Inf2) instances provided cost-effective LLM inference, reducing infrastructure costs by 40% compared to GPU-based models.
  • Amazon Cognito was integrated for secure role-based authentication, ensuring financial advisors had controlled access to investment insights.

Benefits

The AI-powered investment assistant delivered significant efficiency, accuracy, and cost advantages:
Faster Investment Research & Risk Analysis
  • Reduced investment research time by 70%, allowing advisors to retrieve key insights within seconds instead of hours.
  • AI-generated fund comparisons and performance summaries cut manual data analysis by 50%.
Highly Accurate, Document-Based Insights
  • AI-generated responses were strictly document-backed, ensuring credible, audit-ready investment research.
  • BGE-M3 embeddings + Pinecone vector search enabled more relevant, context-aware investment insights.
Cost-Optimized AI Deployment
  • Using AWS Inferentia2 (Inf2) instead of GPU-based instances, the firm reduced AI inference costs by 40%.
  • The pre-built AI retrieval framework from ekSource accelerated deployment, cutting costs by 35-40% compared to industry alternatives.
Scalable, Secure AI Infrastructure
  • The self-hosted Llama 3.1 model + BGE-M3 embeddings ensured full compliance with financial security policies.
  • The RAG pipeline was future-proof, supporting scalability across additional investment data sources.
This PoC successfully demonstrated how an AI-powered investment intelligence assistant can revolutionize financial research and portfolio management. The solution provided an intuitive, scalable, and cost-effective way for portfolio managers to access critical investment insights in real-time, reducing search time and enhancing decision-making.
With this strong foundation, the firm can now explore:
  • Scaling AI-driven investment research across additional financial datasets.
  • Integrating real-time stock feeds and predictive analytics for portfolio modeling.
  • Expanding the AI-powered assistant to support institutional investors and wealth management teams.
By leveraging ekSource’s AI expertise, proprietary document retrieval frameworks, and deep knowledge of financial AI solutions, the client achieved a highly secure, AI-powered investment research assistant at a fraction of the cost and timeline compared to competitors.
ekSource continues to help financial institutions build AI solutions that drive efficiency, automation, and competitive advantage in wealth management.