Project Tokyo

• 

  Cost and Anomalies graphs and horizon selection

• 

  Home Page to interact with Agentic AI

• 

  LLM driven Rightsizing Recommendations - Part 2

• 

  End-of-month and End-of-quarter forecasting with horizon selection

• 

  LLM driven Rightsizing Recommendations

• 

  Architecture Diagram

Inspiration

Cloud bills can change overnight — leaving teams guessing about the “why” behind every spike. We wanted to shift from reactive cost analysis to proactive, AI-driven FinOps intelligence.

What it does

Agentic FinOps detects anomalies in spend, forecasts upcoming costs, compares pricing across AWS, Azure, and GCP, and generates actionable savings insights — all through an interactive Streamlit dashboard.

How we built it

We ingested AWS CUR data into S3, queried it via Athena, and orchestrated insights with Lambda and Bedrock AgentCore for LLM-driven reasoning. The Streamlit UI provides unified visibility for anomalies, forecasts, and cross-cloud cost comparisons.

Challenges we ran into

Unifying pricing models across clouds, ensuring accurate anomaly thresholds, and aligning Lambda–Bedrock responses with strict OpenAPI schemas inside a sandboxed environment.

Accomplishments that we're proud of

Delivered a complete, working multi-cloud FinOps agent — integrating cost detection, forecasting, and optimization through AWS native services and an intelligent AI layer.

What we learned

How to blend LLM reasoning with FinOps automation, structure data pipelines for transparency, and make complex cloud billing insights explainable to business users.

What's next for Agentic FinOps

Integrate with real-time budgets and alerts, enable autonomous optimization actions, and expand forecasting and ROI analytics across multi-account, multi-cloud setups.

Built With

• athena
• bedrock
• lambda
• python
• s3
• streamlit