CTA - Snowflake updates
Date: February 6, 2026 Source: Granola Meeting ID: c9567ffe-cbbe-445f-af12-f6ae5a502ad1 URL: https://notes.granola.ai/t/c9567ffe-cbbe-445f-af12-f6ae5a502ad1
Participants:
- Uttam Kumaran (Brainforge)
- Holly Condos (Brainforge)
- Awaish Kumar (Brainforge)
Summary
CTA Project Status & Snowflake Implementation
- Current progress: 1/3 of data sources landed in Snowflake (60-70 total sources)
- Prioritizing P0 sources post-CES
- Using Python REST APIs, Polyatomics ETL tool, and Snowflake shares
- dbt structure: RAW → intermediate → mart schemas
- Stakeholder: Katherine Bayless (technical champion)
- Evaluating move from Power BI to Snowflake-native reporting
- Interested in Streamlit + Cortex natural language capabilities
- Has existing budget, expects >$15K annual contract
- Previously paid $30K for Redshift doing 1/4 of planned Snowflake workloads
- Reporting scope covers entire CTA operations
- Membership reporting, sponsorships, events registration, badge scans
- Many stakeholders currently have no reporting or rely on manual exports
- Moving from legacy Power BI + on-prem Postgres to modern data stack
Snowflake AI & Natural Language Strategy
- Goal: Replace traditional BI dashboards with natural language querying
- Building “golden dataset” with easy/medium/hard test questions
- Focus on complex multi-table joins vs simple “how many orders today” queries
- Need semantic layer + observability for query monitoring and evaluation
- Technical requirements for Brainforge rollout
- Query observability: see what users ask, query times, evaluation metrics
- Semantic layer guidance for LLM table selection
- Training resources for team dissemination across 4-5 client implementations
- Snowflake Cortex Code CLI capabilities
- Context-aware agent for Snowflake environments
- Can build production pipelines and semantic views from specifications
- Supports custom instructions and orchestration
Next Steps
- Matt & Vince: Collaborate on pricing document this month
- Include traditional credits + new AI features (Streamlit, Cortex)
- Multi-year cost projection for CTA planning
- Vince: Send architecture documentation and reference materials
- Focus on semantic layer creation, Cortex Code CLI, Streamlit+LLM integration
- Skip basics, prioritize advanced observability and depth features
- Uttam: Present proposal to Katherine once pricing/technical details ready
- Confident in contract closure with proper documentation
- Follow-up via existing Slack thread (adding Vince)