Fraud Detection Pipeline with Data Agent Kit & Antigravity IDE

Welcome to this hands-on technical workshop! Detecting fraudulent transactions in modern enterprise environments requires real-time intelligence, clean data pipelines, and fast cross-service investigations. In this session, we will build a production-grade Fraud Detection Pipeline using Google's Data Agent Kit inside the Antigravity IDE to discover suspicious patterns, orchestrate robust transformation models, and deploy automated predictive defenses.

You will learn how to leverage agentic tooling and Model Context Protocol (MCP) integrations to trace anomalies across operational databases, data warehouses, and object storage in a unified workspace.

🏗️ Today’s Technical Roadmap

We will step through building an end-to-end fraud detection and mitigation pipeline:

1. Unified Data Ingestion & Asset Discovery: Locate and catalog transactional data, customer risk profiles, and historical chargebacks across BigQuery, Cloud SQL, and Cloud Storage using the Dataplex Knowledge Catalog directly inside the Antigravity IDE.

2. Cross-Service Fraud Investigation via MCP Tools: Correlate real-time transactional logs with historical fraud patterns in a single conversational session, executing federated queries across distinct storage backends using Model Context Protocol (MCP) tools.

3. Modular dbt Transformation & Feature Engineering: Direct the agent to scaffold a modular dbt (data build tool) project to clean transaction streams, aggregate velocity metrics (e.g., transaction frequency, geographic hops), and run automated schema and data quality tests.

4. Autonomous Anomaly Diagnosis & Pipeline Hardening: Walk through a live debugging scenario where the agent analyzes edge cases (like false-positive spikes or join duplicates), self-diagnoses logic errors, and refactors transformation SQL in real time.

5. Predictive Scoring & Automated Alerting: Apply built-in BigQuery ML anomaly detection and classification models to generate real-time risk scores and construct an automated alerting report for fraud operations teams.

🌟 Why This Session is Critical

• Agentic Data Engineering for Security: Move beyond static query templates. Learn how agent workflows orchestrate multi-database investigations to uncover sophisticated fraud schemes faster.

• Resilient, Tested Feature Stores: Build production-ready dbt models with automated testing to guarantee that mission-critical risk features remain reliable and accurate.

• In-Warehouse Machine Learning: Train and execute fraud classification models directly inside BigQuery, avoiding the latency and security overhead of exporting sensitive financial data.

• High-Impact Portfolio Asset: Delivering an end-to-end fraud detection architecture combining cross-cloud discovery, dbt pipelines, automated debugging, and ML scoring demonstrates elite data and cloud engineering capability.

📝 What You Need to Bring

• Your Laptop (fully charged).

• Antigravity IDE Installed: Configured in your local development workspace.

• A Google Cloud Project: With billing enabled to access BigQuery, Cloud Storage, Cloud SQL, and Dataplex.

• A GitHub Account: Ready to commit and version-control your fraud detection dbt models and SQL scripts.

📚 Essential Developer Resources

• BigQuery ML Anomaly Detection: Detecting Anomalies with BigQuery ML

• Dataplex Knowledge Catalog: Data Discovery & Lineage on Google Cloud

• dbt with BigQuery: Transforming and Testing Data in BigQuery