Shelpuk & Company
    All case studies
    AI Development
    RAG
    Generative AI
    Banking & Finance

    Major Canadian Bank

    RAG Chatbot for Major Canadian Bank

    Built a production-ready Retrieval-Augmented Generation (RAG) system enabling intelligent question-answering across 3,200+ internal documents using Azure OpenAI and LangChain.

    CanadaJan 2025 – Mar 2025

    The Challenge

    The bank needed to enable employees and stakeholders to quickly find accurate, context-aware answers from a vast corpus of internal documents spanning multiple business areas. Manual search was slow and often returned incomplete or irrelevant results.

    Our Solution

    We designed and delivered a complete RAG system consisting of a Python library with DocumentManager and ConversationManager components, a containerized Flask microservice with RESTful APIs, and a data loading utility for ingesting 3,200+ documents. The system supports multi-turn conversations, image-aware document processing, and multiple conversation modes (auto, context-only, general-only).

    The Impact

    • 3,200+ documents ingested and searchable via semantic search
    • Multi-turn conversation support with context-aware responses
    • Image analysis for graphs and visual elements in documents
    • Production-ready containerized microservice with Docker deployment

    Technologies Used

    Azure OpenAI
    LangChain
    PGVector
    Flask
    Docker
    Python
    Ragas

    Related Case Studies

    DeepTrait

    NGS Data Analysis Platform for Drug Discovery

    Co-founded and built an AI-powered NGS data analysis platform that enables lab scientists to discover novel therapeutic targets and diagnostic markers in under six hours — with no coding, scripts, or command lines required.

    Read case study

    AstraZeneca

    AI Product for Medical Affairs

    Developed an AI insights analytics tool and insurance flow pipeline for a global pharmaceutical company, automating CRM record analysis to deliver actionable insights from large-scale Medical Affairs operations.

    Read case study

    Ready to achieve similar results?

    Every engagement begins with a conversation. Let's discuss how AI can drive outcomes for your organization.

    Frequently Asked Questions

    Answers to common questions about this topic.

    Which challenge is this case study solving for Major Canadian Bank?

    Major Canadian Bank needed to solve this challenge: The bank needed to enable employees and stakeholders to quickly find accurate, context-aware answers from a vast corpus of internal documents spanning multiple business areas. Manual search was slow and often returned incomplete or irrelevant results. Shelpuk AI Technology Consulting led the technical direction and implementation discipline in this engagement.

    What delivery approach was used to address that challenge?

    The delivery approach combined AI Development, architecture design, iterative implementation, and technologies such as Azure OpenAI, LangChain, PGVector, Flask, Docker, Python, Ragas to reach operational goals. Shelpuk AI Technology Consulting uses production-oriented delivery methods to move from prototype to operational value.

    What measurable outcomes are highlighted for Major Canadian Bank?

    Key measurable outcomes included: 3,200+ documents ingested and searchable via semantic search; Multi-turn conversation support with context-aware responses; Image analysis for graphs and visual elements in documents; Production-ready containerized microservice with Docker deployment. Shelpuk AI Technology Consulting focuses on KPI-level outcomes and operational reliability in comparable environments.

    Which capabilities and technologies were critical in this project?

    Core capabilities in this project included AI Development supported by technologies such as Azure OpenAI, LangChain, PGVector, Flask, Docker, Python, Ragas. Shelpuk AI Technology Consulting combines platform engineering and AI expertise to deliver maintainable systems.

    How can another organization adapt this approach safely?

    Replication starts with discovery, data readiness checks, compliance alignment, and phased rollout planning. Shelpuk AI Technology Consulting helps organizations tailor this pattern to their domain constraints and governance needs.

    What engagement model is best for similar initiatives?

    Engagement can begin with advisory, a fixed-scope pilot, or a dedicated engineering team depending on timeline and risk. Shelpuk AI Technology Consulting offers consulting, delivery, and enablement paths for similar business goals.