loader image

Kitsilano Technologies Limited

Menu
Get in Touch
Menu
Get in Touch

Mizizi AI Delivers Intelligent Knowledge Retrieval with Generative AI on AWS

Case Study Title:

Mizizi AI Delivers Intelligent Knowledge Retrieval with Generative AI on AWS

About the Customer:

Mizizi AI is an emerging AI-driven knowledge solutions company focused on building intelligent assistants for enterprise environments. The organization supports businesses by offering chatbot-based solutions that help internal teams and customer service staff access critical information with ease. Mizizi operates across East Africa and partners with public institutions, startups, and SMEs to deploy scalable AI products.

Customer Challenge:

As Mizizi scaled its AI knowledge solutions for enterprise customers, it faced several challenges around performance, data governance, and scalability:

  1. Fragmented Knowledge Sources
    Enterprises often maintained policies, training manuals, and SOPs in disparate formats and platforms. Mizizi needed to unify these into a centralized knowledge base accessible via intelligent search.

  2. Slow Query Response Times
    Without semantic search and vector-based retrieval, answering user queries using traditional keyword-based search was inefficient and often inaccurate.

  3. Manual Query Response Processes
    Human agents were involved in answering frequently asked questions. This limited scalability and was prone to delays and inconsistent responses.

  4. Scalability and Security Concerns
    With growing data volumes and usage spikes, the solution needed to ensure low-latency performance, high availability, and secure data handling within enterprise-grade guardrails.

Risks if Unaddressed:

  • Poor user satisfaction

  • High operational costs for human-led support

  • Inaccurate information retrieval

  • Weak adoption of AI knowledge assistants

Partner Solution:

To address these issues, Kitsilano Technologies Limited, an AWS Advanced Consulting Partner and specialist in Generative AI, implemented a robust Retrieval-Augmented Generation (RAG) architecture hosted entirely on AWS infrastructure.

✨ Architecture Overview:

Mizizi AI’s deployed solution integrates:

  • Vector-based document search using Amazon OpenSearch

  • Preprocessing and embedding generation using Amazon SageMaker

  • Real-time, natural language response generation with Amazon Bedrock

  • Secure and scalable hosting using Amazon VPC, ALB, and S3

🔐 1. User Access and Query Handling

  • Users interact via a custom Chatbot UI, accessible via desktop and mobile.

  • The UI passes queries to a public subnet-hosted Web Server.

  • The Amazon OpenSearch Service (configured as a Vector Database) intermediates between user queries and the backend processing engine, ensuring low-latency semantic search.

⚙️ 2. Compute and Knowledge Retrieval Layer

  • The Web Server, hosted securely in a VPC, orchestrates flow between the chatbot, OpenSearch, and backend AI services.

  • Amazon Bedrock serves two roles:

    • The Foundation Model processes user questions and context to generate accurate, natural responses.

    • The Knowledge Base within Bedrock supports contextual awareness by connecting to internal documents and repositories.

🧠 3. Embedding and ML Processing

  • Amazon SageMaker plays a central role in machine learning operations:

    • SageMaker LLM for Embedding transforms enterprise corpus documents into dense vector embeddings.

    • SageMaker Processing Jobs clean, tokenize, and process raw documents stored in Amazon S3.

    • SageMaker LLM for Text Generation supports enhanced answer generation workflows.

🗂️ 4. Corpus Storage and Management

  • Enterprise documents and internal knowledge (PDFs, DOCX, manuals) are stored in Amazon S3.

  • These are ingested into the ML workflow for preprocessing and vectorization, making them accessible through semantic search and LLM-generated responses.

Results and Benefits:

The Mizizi AI solution has been deployed to enterprise and institutional customers with resounding success.

📊 Measurable Results:

  • 💬 80–90% of queries auto-resolved via chatbot with no human intervention

  • 🕒 70% reduction in average response time for knowledge queries

  • 📉 50% reduction in manual support load, freeing up human agents for complex tasks

  • 🔐 Full compliance with enterprise data protection policies using VPC, IAM, and KMS

🎯 Strategic Advantages:

  • Intelligent AI knowledge access powered by Generative AI

  • Scalability to serve thousands of users concurrently without performance drop

  • Seamless integration with future AI workloads using Bedrock and SageMaker

  • Demonstrated cost savings by reducing human support overhead

About the Partner:

Kitsilano Technologies Limited is an Advanced AWS Partner based in Kenya with verified Service Delivery Programs for Amazon RDS, Glue, Systems Manager, and QuickSight. Kitsilano is the only East African AWS partner with the SAP Competency and is at the forefront of deploying Generative AI and intelligent enterprise automation solutions. Kitsilano helps businesses modernize using AI/ML, serverless architectures, and enterprise-ready AWS services.

✅ Case Study Summary

  • Customer: Mizizi AI (AI Solutions Provider, East Africa)

  • Challenge: Unifying and accessing enterprise knowledge efficiently using AI

  • Solution: Generative AI-powered chatbot using RAG with OpenSearch, SageMaker, Bedrock

  • AWS Services: Amazon Bedrock, SageMaker, OpenSearch, S3, VPC

  • Outcomes: 80%+ auto-resolved queries, reduced support burden, fast response times

  • Partner: Kitsilano Technologies Limited (Advanced AWS Partner, SAP Competency Holder)