Beyond the Balance Sheet: Transactional Data as a Window to Company Health
summary
Claims that daily transaction data reveal company health faster than year‑end accounts and that Heartbeat turns that into actionable signals using heuristics and ML. Built a pipeline with anonymization, cloud preprocessing and model weights deployed into the app, trained on a multi‑source feed (uploader cites 50 TB, presenter mentions a 10 TB contract). Demo shows a client that failed two months before public default. Product parses bank statements, generates dimensions and alerts, computes a coverage score, and prototypes default models (XGBoost/regression) and a graph database for network effects.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:11Speaker background
- 2:39Traditional reporting limits
- 6:10Why transactional data matters
- 8:21Missed default prediction case
- 11:56Scaling with big data & ML
- 13:26Tool architecture
- 15:58Graph databases & analysis
- 17:44Coverage score & alerts
- 19:29Categorization challenges
- 21:17Q&A
- 21:39Real-time calculations
- 24:11Questions on machine learning models and graph usage
- 26:58Future plans
- 27:43Closing remarks
related talks
25mAnalyzing Daily Bank Transactions: Maximizing Insights
24mRetail Transactions at scale for AI - From PoC to production
10mBuilding High-Performance Financial Data Pipelines
16mGenerative AI for Claims and Fraud
20mSession: From Chaos to Clarity: The Journey to Building a Money-Saving BI Tool
21mCloud-Native Observability for Financial Systems
41mBeyond the Dashboard: Building Trustworthy Data with Observability at PAYBACK
17mOptimizing Data-Driven Decisions