🏔️ Snowflake Asset Management AI Transformation

🎯 Objective: Build a comprehensive AI-powered asset management system that combines document search, portfolio analytics, and intelligent assistance capabilities.

This demo provides detailed step-by-step instructions for setting up a sophisticated Asset Management AI demo in Snowflake, featuring:

• 🔍 Cortex Search - Intelligent document search and retrieval
• 📊 Cortex Analyst - Advanced portfolio data analysis
• 🖥️ Streamlit Integration - Interactive web application
• 🤖 Snowflake Intelligence - AI agent with advanced portfolio analysis and document search

🚀 What You'll Build

By the end of this setup, you'll have a fully functional AI system that can:

✅ Search through research documents using natural language
✅ Analyze portfolio data with conversational queries
✅ Provide intelligent recommendations combining both data sources
✅ Deliver insights through an intuitive web interface

📋 Repository Structure

├── README.md                             # This comprehensive guide
├── LEGAL.md                              # Legal notices
├── LICENSE                               # License information
├── notebooks/                            # Snowflake Notebooks
│   ├── environment.yml                   # Python package dependencies
│   └── 0_start_here.ipynb                # Main setup notebook for Cortex Search
├── scripts/                              # SQL and configuration files
│   ├── setup.sql                         # Database and schema setup script
│   └── semantic_models/                  # Sample semantic models for Cortex Analyst
│       └── PORTFOLIO_ANALYSIS.yaml       # yaml model for Cortex Analyst
│   └── generated_pdfs/                   # Sample research documents
│       └── *.pdf                         # PDF research documents for demo
└── streamlit/                            # Streamlit application
    ├── app.py                            # Streamlit dashboard code

📋 Prerequisites

🔑 Required Access & Permissions

☑️ Snowflake account with appropriate privileges
☑️ SQL execution permissions
☑️ Ability to create databases, schemas, and stages
☑️ Access to Snowflake Notebooks
☑️ Access to Snowflake Intelligence features
☑️ Cross Region Cortex Inference enabled (if required for your region)

📁 Required Demo Files

Ensure you have all the following files ready:

File Type	Description	Location	Usage
📄 SQL setup script	Database and schema creation	scripts/setup.sql	Step 1
📚 PDF documents	Sample research documents (download first)	scripts/generated_pdfs/ folder	Step 2 & 3
📓 Jupyter notebook	Cortex Search setup and processing (download first)	notebooks/0_start_here.ipynb	Step 2 & 4
📊 Cortex Analyst	Portfolio analysis semantic model	scripts/PORTFOLIO_ANALYSIS.yaml	Step 5
🤖 Snowflake Intelligence	AI agent configuration and testing	Step 6	Step 6
🖥️ Streamlit app	Interactive dashboard code	streamlit/app.py	Step 7

---

🏗️ Setup Instructions

1️⃣ Step 1: Execute SQL Setup Statements

1.1 Access Snowsight and Open Worksheets

1. Open Snowflake in your web browser (Snowsight interface)
2. Navigate to Projects → Worksheets in the left sidebar
3. Click + Worksheet to create a new SQL worksheet

1.2 Import and Execute the Setup Script

1. Locate the scripts/setup.sql file from this repository
2. Open the file and copy all contents
3. Paste the SQL statements into your new Snowflake worksheet
4. Important: The script uses accountadmin role for initial setup
5. Execute all statements by clicking ▶ Run All or pressing Ctrl+Shift+Enter

1.3 What the Setup Script Creates

The enhanced setup.sql script will automatically create:

Security & Access:

• Role: asset_management_ai_role with Cortex and Streamlit capabilities
• Warehouse: asset_management_ai_wh (Small, auto-suspend after 60 seconds)
• Grants: Comprehensive permissions for AI services and data access

Data Infrastructure:

• Database: ASSET_MANAGEMENT_AI
• Schema: RESEARCH_ANALYTICS
• Stage: RESEARCH_DOCS internal stage (with encryption enabled)
• Table: PORTFOLIO_HOLDINGS with sample portfolio data
• Sample Data: 13 portfolio holdings across Technology, Healthcare, ESG, Real Estate, and Utilities sectors

Snowflake Intelligence Setup:

• Database: snowflake_intelligence for AI agent storage
• Schema: snowflake_intelligence.agents for agent management
• Permissions: CREATE AGENT privileges for the demo role

Note: If you continue with accountadmin, that's fine too - the setup script grants all necessary permissions to both roles.

---

2️⃣ Step 2: Download Demo Files

2.1 Download the Notebook

1. Navigate to the notebooks/ folder in this repository
2. Download the notebook file: 0_start_here.ipynb
   • Click on the file in GitHub
   • Click the Download button
   • Alternative: Clone the entire repository to get all files locally
3. Note the location where you saved the notebook file for Step 4

2.2 Download Research Documents

1. Navigate to the scripts/generated_pdfs/ folder in this repository
2. Download all PDF files individually or clone the repository to get them locally
3. Ensure you have all 8 research documents downloaded:
   • NovaTech Insights - Technology Sector Overview.pdf
   • NovaTech Insights - Emerging Markets.pdf
   • NovaTech Insights - ESG Trends.pdf
   • Gaming Industry Trends in 2025.pdf
   • FinTech Innovation Investment Opportunities.pdf
   • Dissertation on the Utility Sector.pdf
   • Stellar Motors Inc. - 10-K Annual Report.pdf
   • SNOW_2025.pdf

---

3️⃣ Step 3: Upload Research Documents

3.1 Locate Your Downloaded Research Documents

1. Find the downloaded PDF files from Step 2.2 on your local machine
2. These documents are specifically chosen to match the portfolio holdings created in Step 1
3. Document types include:
   • Technology Sector Analysis: NovaTech Insights - Technology Sector Overview.pdf
   • Emerging Markets Research: NovaTech Insights - Emerging Markets.pdf
   • ESG/Sustainability Analysis: NovaTech Insights - ESG Trends.pdf
   • Gaming Industry Trends: Gaming Industry Trends in 2025.pdf
   • FinTech Investment Research: FinTech Innovation Investment Opportunities.pdf
   • Utility Sector Analysis: Dissertation on the Utility Sector.pdf
   • Company Annual Report: Stellar Motors Inc. - 10-K Annual Report.pdf
   • Market Analysis: SNOW_2025.pdf

3.2 Upload Documents via Snowsight Data Interface

Method A: Snowsight Web Interface (Recommended)

1. In Snowsight, navigate to Data → Databases
2. Expand ASSET_MANAGEMENT_AI → RESEARCH_ANALYTICS
3. Click on Stages and select the RESEARCH_DOCS stage
4. Click + Files button to open the upload dialog
5. Select all 8 downloaded PDF files from your local machine:
   • Browse to where you saved the files from Step 2.2
   • You can drag and drop multiple files at once
   • Or use Ctrl+Click (Windows) / Cmd+Click (Mac) to select multiple files
6. Click Upload to transfer all documents to the stage
7. Wait for the upload progress to complete

Method B: SQL Commands (Alternative) If you prefer using SQL commands, you can upload via PUT statements:

-- Ensure you're in the correct context
USE DATABASE ASSET_MANAGEMENT_AI;
USE SCHEMA RESEARCH_ANALYTICS;

-- Upload files (replace with actual local file paths where you downloaded them)
PUT file:///path/to/your/downloads/NovaTech_Insights_Technology_Sector_Overview.pdf @RESEARCH_DOCS;
PUT file:///path/to/your/downloads/Gaming_Industry_Trends_in_2025.pdf @RESEARCH_DOCS;
PUT file:///path/to/your/downloads/Stellar_Motors_Inc_10K_Annual_Report.pdf @RESEARCH_DOCS;
-- Repeat for each of the 8 downloaded PDF files

4️⃣ Step 4: Load Notebook and Create Cortex Search Service

4.1 Access Snowflake Notebooks in Snowsight

1. In Snowsight, navigate to Projects → Notebooks in the left sidebar
2. Click + Notebook to create a new notebook
3. Choose Import .ipynb file option

4.2 Import the Downloaded Notebook

1. Click Browse and select the 0_start_here.ipynb file you downloaded in Step 2.1
2. Configure the notebook environment:
   • Database: ASSET_MANAGEMENT_AI
   • Schema: RESEARCH_ANALYTICS
   • Warehouse: asset_management_ai_wh (created in Step 1) or any S/M size warehouse
3. Click Create Notebook

4.3 Understanding the Notebook Structure

The 0_start_here.ipynb notebook contains:

• Executive Summary: Overview of the AI research pipeline transformation
• Solution Architecture: Detailed explanation of the 5-component system
• Document Processing Pipeline: Code for processing PDF research documents
• Cortex Search Setup: Instructions for creating the INVESTMENT_SEARCH_SVC service
• Interactive Examples: Sample queries and search demonstrations

4.4 Execute Notebook with Uploaded Documents

1. Now that documents are uploaded from Step 3, proceed with document processing
2. Execute cells sequentially from top to bottom, following the notebook's guidance
3. The notebook will establish database connections and process the uploaded documents

4.5 Document Processing Workflow

The notebook will guide you through:

1. Document Parsing: Extract text content from the uploaded PDFs
2. Text Chunking: Break documents into optimal-sized segments for search
3. Embedding Generation: Create vector embeddings using Snowflake Cortex
4. Metadata Extraction: Add relevant tags and classifications
5. Search Service Creation: Deploy the INVESTMENT_SEARCH_SVC Cortex Search service

4.6 Add More Documents (Optional)

If you have additional research documents to add later:

1. Upload them to the RESEARCH_DOCS stage using the same methods from Step 3.2
2. Re-run the document processing cells in the notebook to include new documents
3. The search service will automatically include the new content

---

5️⃣ Step 5: Configure Cortex Analyst

5.1 Access Cortex Analyst

1. Navigate to AI & ML → Cortex Analyst in Snowsight

5.2 Upload Configuration File

1. From the Create new dropdown, select Upload YAML file
2. Download the scripts/PORTFOLIO_ANALYSIS.yaml file from this repository
3. Upload the YAML document from this repo

5.3 Configure Stage and Path

1. Select the stage we created earlier: RESEARCH_DOCS
2. Add the path as: semantic_models
3. Click Upload

5.4 Save Semantic Model

1. Review the configuration to ensure it correctly references your database and schema
2. Click Save to save the new Cortex Analyst semantic model
3. Note the semantic model name for the next step (typically named after your YAML file)

---

6️⃣ Step 6: Create and Test Snowflake Intelligence Agent

6.1 Access Snowflake Intelligence

1. Navigate to AI & ML → Agents in Snowsight
2. Click + Agent to create a new agent

6.2 Configure Agent Basic Settings

Configure your agent with these details:

• Platform Integration: Check Create this agent for Snowflake Intelligence
• Agent Name: Research_Assistant
• Display Name: Research Assistant
• Click Create
• Click Edit and navigate to Orchestration
• Instructions: Copy and paste this instruction set in Orchestration instructions:

You are an expert investment research assistant for Global Asset Management. You have access to:

1. A comprehensive research document library via Cortex Search containing reports on technology, healthcare, ESG, and real estate sectors
2. Portfolio data for analyzing current holdings and performance

Your role is to:
- Answer questions about investment research and market analysis
- Provide insights on portfolio holdings and sector performance
- Search through research documents to find relevant information
- Analyze portfolio data to support investment decisions
- Combine document insights with portfolio data for comprehensive analysis

Always provide data-driven responses and cite relevant sources when possible.

6.3 Add Cortex Analyst Tool

1. From Tools, click + Add Tool again
2. Select Cortex Analyst
3. Select semantic model file: Choose the file we uploaded when creating Cortex Analyst
4. Select database/schema/stage: Choose ASSET_MANAGEMENT_AI → RESEARCH_ANALYTICS → RESEARCH_DOCS and select the file we uploaded (semantic_models/PORTFOLIO_ANALYSIS.yaml)
5. Add the following details:
   • Name: Portfolio_analysis
   • Description: Analyze portfolio data and generate insights or generate via Cortex
   • Timeout: 60 seconds
6. Click Add

6.4 Add Cortex Search Tool

1. Navigate to Tools
2. Click + Add Tool
3. Select Cortex Search
4. Name: Document_search
5. Tool Description: Search through investment research documents and reports
6. Choose ASSET_MANAGEMENT_AI → RESEARCH_ANALYTICS → INVESTMENT_SEARCH_SVC service created in Step 4
7. Click Add

6.5 Finalize and Test Agent

1. Review all settings and tools
2. Ensure both Cortex Search and Cortex Analyst tools are properly configured and click Save
3. Two ways to interact with your agent:
   • Option A: Test directly from the agent configuration screen (chat interface at bottom)
   • Option B: Navigate to Snowsight → AI → Snowflake Intelligence and select your agent
4. Test immediately with these sample queries:

📊 Portfolio Analysis Queries (using Cortex Analyst):

• "What are our current portfolio holdings by sector?"
• "Which holdings have the highest risk levels?"
• "Chart our portfolio holdings by sector and risk"
• "Summarize our investment portfolio and outline the risk associated with each"

🔍 Document Research Queries (using Cortex Search):

• "What are the key technology trends mentioned in our research documents?"
• "Find information about ESG investment opportunities in our research library"
• "Compare the technology sector insights from our different research reports"

💡 Strategic Portfolio Insights (combining both tools):

• "What are the primary growth catalysts mentioned for our Technology holdings?"
• "What ESG trends are likely to impact our portfolio?"
• "Provide a summary of our Technology Sector Overview report and how it will impact our portfolio"

6.6 Demo Capabilities and Access Options

1. Try portfolio analysis queries to see how the agent analyzes your portfolio data
2. Test document search to see how the agent searches through your uploaded research documents
3. Combine both capabilities by asking questions that require both portfolio data and document insights
4. Note the comprehensive responses with relevant citations from both data sources

Access Your Agent:

• In Configuration Screen: Use the chat interface at the bottom of the agent configuration screen
• Snowflake Intelligence Portal: Navigate to Snowsight → AI → Snowflake Intelligence and select your "Research Assistant" agent
• Better Experience: The Snowflake Intelligence portal provides a full-screen chat interface optimized for longer conversations

Your agent can now:

• Analyze portfolio data using natural language queries (Cortex Analyst)
• Search through research documents using semantic search (Cortex Search)
• Generate insights by combining portfolio data with document research
• Provide intelligent recommendations based on both data sources
• Create visualizations and charts for portfolio analysis
• Extract specific information from your research library and portfolio data

---

7️⃣ Step 7: Create Streamlit Application

7.1 Prepare the Streamlit Application Code

1. Locate the streamlit/app.py file from this repository
2. Open the file and copy all the Python code contents
3. This file contains the complete interactive dashboard for portfolio analytics and document search

7.2 Create New Streamlit App

1. In Snowsight, navigate to Projects → Streamlit in the left sidebar
2. Click + Streamlit App
3. Configure the app:
   • Name: GAM Research Assistant App (or similar)
   • Database: ASSET_MANAGEMENT_AI
   • Schema: RESEARCH_ANALYTICS
   • Warehouse: asset_management_ai_wh (or same as used for notebook)

7.3 Deploy Streamlit Code

1. Delete any default code in the Streamlit editor
2. Paste the copied Python code from app.py
3. Add plotly from the packages dropdown
4. Click Run to deploy the app

---

🎯 Demo Story: Asset Management AI Transformation

This demonstration showcases how Global Asset Management (GAM) revolutionized their investment research and portfolio management processes using Snowflake's integrated AI platform.

📊 The Challenge

GAM's analysts were spending 30-40% of their time organizing and finding documents instead of analyzing investments:

• Manual classification of research reports and SEC filings
• Keyword-based searches missing critical insights
• Siloed data across different systems preventing comprehensive analysis
• Hours to answer urgent research questions during volatile market conditions

🚀 The Transformation

GAM deployed a comprehensive AI research platform that delivers:

• ⚡ 30-second insights instead of hours of manual research
• 🎯 Semantic understanding of investment concepts and relationships
• 📊 Real-time portfolio analysis with AI-driven risk assessment
• 🤖 Conversational research assistant for instant answers

💡 Why Snowflake Makes This Possible

• 🚀 Integrated AI Stack: LLMs, embeddings, and vector search natively within the data warehouse
• ⚡ No Infrastructure Management: Automatic scaling without GPU cluster management
• 💰 Elastic Economics: Pay only for actual compute usage
• 🔒 Enterprise Security: SOC 2 compliance and encryption built-in
• 📊 Data Gravity: Research documents and portfolio data in the same platform

📈 Business Impact

• 90% reduction in research organization time
• Instant access to insights across thousands of documents
• Real-time analysis enabling rapid response to market changes
• Zero infrastructure management reducing IT overhead

---

🆘 Support

• 💬 Snowflake Community
• 📖 Snowflake Documentation
• 🤖 Cortex AI Documentation
• 🔍 Cortex Search Documentation
• 🧠 Snowflake Intelligence Documentation