Had a chance to attend an event “Data for Breakfast” organized by Snowflake SG. Here are some notes from the event

Address by Mike Scarpeli CFO Snowflake

Data sharing
9500 customers
2400 listings on marketplace
461 - $1M customers
ASEAN - 130 customers
9437 customers in Q4 FY2024
691 customers Fortune 200
7000 FTE
Pricing gets cheaper as time goes
Workloads
- AI/ML
- Applications
- Cybersecurity
- Data Engineering
- Datawarehouse
- Datalake
- Unistore

Build a Data Foundation with the Snowflake Data Cloud

There is no AI strategy without a data strategy
Simplify your data foundation : 5 design principles for your data foundation
- Endless silos to Unified data
- Multiple products and services for different workloads to One platform all workloads
  - Snowpark
  - Data Lake
  - Data Lakehouse
  - Data warehouse
  - Date Mesh / Fabric
  - Average daily queries - 3.9 B
- Piecemeal policies to Universal governance
  - Compliance
  - Security
  - Privacy
  - Access
  - Interoperability
- Hidden costs to Optimal TCO
- Limited to internal data only to 360 view with external data, data collaboration and data monetization
- More than 2000 dataset, models, Snowflake native apps
- More than 530 Providers on the marketplace
Accelerate AI
- Long cycles + AI limited to experts to AI in seconds, easy for everyone
  - Single platform for end to end ML
  - Notebooks
  - Feature store
  - Snowpark ML Modeling
  - Snowpark Model Registry
  - Streamlit in Snowflake
- Platform Overview
  - Use AI in seconds
    - Document AI
    - Universal Search
    - Snowflake Copilot
  - Apps in Minutes
    - Streamlit
  - Fully Custom in hours
    - Custom UI
    - Custom Orchestration
- Snowflake Cortex: Models and Search
  - Use complete, embed, vector search and more
  - Serverless AI and LLM functions
- Snowflake container services
  - Bring your own OSS LLMs
  - Make it your own Fine tuning
  - Use Marketplace - Partner LLMs
  - CSP LLMs
  - Azure OpenAI
  - AWS Bedrock
- Snowflake - Governed Data and Models
- Time and resources on managing infrastructure to Build custom AI apps in minutes
- Experiments mean moving data, leading to risk to AI/ML on enterprise data within Snowflake’s security boundary
Scale with applications
- Architectural Complexity to Build any app with ease
- Unpredictable operational burden to Efficient scaling, given apps are the future consumption layer
- Limited distribution to Secure, global deployment to run apps where data already is
- Faster time to market + Lower TCO
Customers
- StateStreet full application on Snowflake
Demo
- Address customer calls in near real-time and refresh the data every minute with Dynamic tables
- Leverage Snowflake Cortex for Generative AI
- Build an Streamlit app
Snowflake notebooks in Private Preview
- Notebook support using Snowpark
Snowflake Cortex
- LLMs as a managed Service
- Functions given to call various LLMs
  - snowflake.cortex.sentiment
Snowpipe streaming
- Dynamic tables where sentiment processing is done real time
- DYNAMIC TABLE construct to create dynamic tables
- UI to create all the connections with Dynamic tables in Snowsight UI
Streamlit app
- Embed streamlit apps in the platform
- Managed hosting of an app
- No need to separately host this app on AWS
  - Share with in the Snowflake platform
- Have built some forecast function

Build the future of insights and AI with Snowflake and AWS

Laying the foundation of the “insights driven” organization in an AI powered world
- By 2026, global spending on AI will reach 300 billion usd growing 4.2 times faster than average IT spend
- AI across several business units
- Enhance customer experiences
- Boost employee productivity
- Optimize business processes
- Your data is the differentiator
How can AWS and Snowflake power your Modern Data platform journey
Join Customer case study - Human managed
- AI powered data platform based out of SG
- AI-powered data platform for businesses to make smarter and faster decisions for cyber, digital and risk outcomes
  - Collect, process and store data from any source
  - Apply conditions, rules and models for use cases
  - Deliver intelligence, decisions and actions
  - Operationalize intel and improve models

Partners mentioned during the event

cloudmile
- system integrator based in Taiwan, HK, Vietnam, Malaysia, Singapore, Indonesia and Phillipines
blazeclan
black diamond
braze
cloudvalley
fivetran
dbtlabs
izeno

Snowflake Genius bar demo

Saw a nice demo where Streamlit apps can easily be built within the Snowflake platform and can be shared via MarketPlace

Hands-on-Labs: Automating Data Pipelines to Power your ML Journey

User guide
Quick Start Guide
Unique architecture as a Platform
- Optimized Storage
  - Adaptive Caching
  - Zero Copy Clone
  - Time Travel
- Elastic Performance engine
  - Scale up/Down
  - ELT
  - Data Science
- Cloud Service
  - Software as a service
  - optimization
  - Management
  - transactions
  - Security and Governance
Fivetran
- Founded in 2012
- Made access to data reliable
- Make access to data as simple and reliably
- Fully managed, hybrid and self-hosted architecture
- 400+ pre-built connectors
- Connectors set up in give minutes
Differentiators
- Automatic Data Updates
- Automatic Schema Replication
- Automated Recovery from Failure
- Micro-batched architecture
- Slowly Changing, Type 2 Diension Data
- Extensibility beyond our 300+ sources
- Efficient writes to destination to lower compute costs
OpenAI is the customer of Fivetran
- Need to delete chat logs at customer request
- Centralize chat data from Cosmos DB in enterprise data warehouse
- Replication of all the data in snowflake databricks