RBNA · BI, Analytics & Data Engineering
AI Build Day
11 builders. 4 hours to ship — real BI, analytics, and data engineering products, built with AI and working in production.
The challenge
The tracks
- · DAAP Flow + BEx Query KPIs in Semantic View
- · Full End-to-End Data Product with Cortex Skills
- · Build Your First Agent End-to-End with Skills
- · Own workflow
Judging criteria
Tools in play
Swipe, scroll, or use ← → to see the eleven builds.
Slot 1 of 11
Uma
DAAP Flow + BEx Query KPIs in Semantic View
What they built
Snowflake semantic view with BW KPI logic. Creating an Order Profile within EWM and linking with EWM tasks and status.
Tools
Impact
Faster semantic build, fewer manual translation errors, consistent KPI definitions.
Why this matters
Proves the BW → Cortex pattern is repeatable across reports — a template for the rest of the BW estate.
Judges' ratings
Praised by 5 judges
An impressive first-time build — a full semantic view and agent stood up on the fly, laying down a repeatable BW/SF migration pattern.
Slot 2 of 11
Shaifali
Full End-to-End Data Product with Cortex Skills
What they built
Cortex Skill that profiles any Snowflake table and outputs a production-readiness report: governance, DQ, lineage (dbt + pipelines), join safety, usage, recommendations. Generates 8 sub-skills, orchestrator, deploy scripts, docs.
Tools
Impact
Per analysis: 165 min → 2 min (~98% faster). Catches join fan-outs and DQ issues before they ship as bugs.
Why this matters
One command, any table, any schema — instant onboarding for every engineer. Highest leverage skill of the day.
Judges' ratings
Praised by 5 judges
An intuitive, every-day-use tool that nails the problem framing and unlocks real time savings across the team.
Slot 3 of 11
Avi
Own workflow
What they built
Internal "AI Builders Hub" — a Teams-integrated space to post and browse AI ideas, prompts, demos, and questions.
Tools
Impact
Working POC in a couple of hours vs. an estimated 6–8 weeks built manually (~10× acceleration). Eliminates redundant Q&A and the chat-thread archaeology problem — team contributes and finds AI solutions in one place.
Why this matters
Centralizes AI solutions across the org — best practices become readily available, prevents duplicate work across teams. Hub itself is the reusable artifact.
Judges' ratings
Praised by 5 judges
A creative, end-to-end AI play that drives company-wide adoption by uniting creators and early adopters.
Slot 4 of 11
Luis
DAAP Flow + BEx Query KPIs in Semantic View
What they built
Semantic view for the RBNA MPR — PSR/WvC report in Snowflake, replicating the full KPI logic from BW (calculated and restricted key figures, exception aggregation), structured as Facts and Metrics. Started from Confluence documentation, exported to Excel, adapted so Cortex could generate the Facts and Metrics aligned to BW definitions.
Tools
Impact
49 Facts and 70 Metrics built. Cortex accelerated the build phase materially, improved consistency in KPI definitions, and reduced human error when translating complex BW logic.
Why this matters
Approach (Confluence → Excel → Cortex-generated Facts/Metrics) generalizes to any BW report that needs to land in Snowflake — Uma is using the same pattern for EWM.
Judges' ratings
Praised by 5 judges
Solid porting of legacy BW logic into Cortex — a strong proof of concept delivering immediate business value on a report everyone uses.
Slot 5 of 11
Ernesto
Own workflow
What they built
Monitoring and Auto-Healing Agent for daily ingestion jobs into DCP.
Tools
Impact
Shifts daily ingestion ops from alert-driven manual fixes to autonomous self-healing. Major reduction in on-call toil.
Why this matters
Closed-loop ops: detects, diagnoses, and fixes itself. The dream of every data engineer.
Judges' ratings
Praised by 5 judges
Pipelines that fix themselves — a closed-loop, self-healing system that brings monitoring front and center.
Slot 6 of 11
Yogesh
Full End-to-End Data Product with Cortex Skills
What they built
Semantic view for SAP TM data on Snowflake.
Tools
Impact
Faster semantic build, fewer manual translation errors, consistent KPI definitions.
Why this matters
Proves the BW → Cortex pattern is repeatable across reports — a template for the rest of the BW estate.
Judges' ratings
Praised by 5 judges
Self-serve analytics on SAP TM data — a real unlock for logistics, built fast without sacrificing quality.
Slot 7 of 11
Harsha
Build Your First Agent End-to-End with Skills
What they built
Data product for fully automated SFTP → Snowflake ingestion, plus a packaged skill markdown for reuse.
Tools
Impact
Productizes a common ingestion pattern. Future SFTP sources onboard in minutes, not days.
Why this matters
Builds the skill alongside the product — knowledge ships as an asset, not a Slack thread.
Judges' ratings
Praised by 5 judges
A reusable skill that onboards future SFTP sources in minutes — interesting, extensible, and ready to scale.
Slot 8 of 11
Apurva
Build Your First Agent End-to-End with Skills
What they built
AI framework to monitor and resolve Snowflake source REFRESH task failures: auto-detects failed tasks, analyzes logs and query history for root cause, and suggests or applies safe fixes for common failure patterns. • SCoco - error analysis, fix suggestions • Claude - failure triage, troubleshooting steps • GitHub Copilot - dev velocity
Tools
Impact
Replaces the email → manual investigation → manual fix loop with automated detection and resolution. Consistent RCA, less reactive firefighting, fewer late-night pages.
Why this matters
First-time agent builder going straight at a daily ops pain point — and a clean blueprint other DEs can clone for their own task suites.
Judges' ratings
Praised by 5 judges
Consistent, automated RCA that tackles pipeline failure head-on — a major win for reliability and resilience.
Slot 9 of 11
Alex
Own workflow
What they built
Skill-ifying the MySQL → Snowflake migration playbook.
Tools
Impact
Per migration: 2 hours → 30 min (optimized) → 2 minutes (skillified). Runs ~8× a year — saves a full work week annually on a single workflow.
Why this matters
Real, measured 60× speedup. Textbook example of turning tribal knowledge into a reusable skill.
Judges' ratings
Praised by 5 judges
Concrete, measurable time savings — a week back per year from a single workflow.
Slot 10 of 11
Deiby
Own workflow
What they built
AI orchestration agent that auto-provisions DCP Data Products — generates YAMLs, technical templates, and publishes docs to Confluence end-to-end.
Tools
Impact
Replaces manual form-filling with a guided dialogue. Zero documentation debt — specs always in sync with config. Library compounds: agent gets smarter as more products ship.
Why this matters
Fully local LLM for data privacy + a self-improving knowledge base. Most ambitious agentic stack of the day.
Judges' ratings
Praised by 5 judges
An ambitious, end-to-end build that pushes AI into the infra layer — self-improving, open-source-powered, and cost-conscious.
Slot 11 of 11
Sam
Own workflow
What they built
Self Improving Wingman Agent: Enabling automatic updating of the Semantic View: Updates based on user questions, creating verified queries and optimizes SQL to use lower compute.
Tools
Impact
Majority of time is spent on improving agents. Automating improvements significanly enhances agent quality while optimizing cost.
Why this matters
Replicate across all agents we build.
Judges' ratings
Praised by 5 judges
AI doing the quality work we never have time for — continuously validating user questions against Snowflake to scale the Wingman journey.