Engineering Athena: Financial Platform at J.P. Morgan
summary
Claims a migration from thousands to ~1,000,000 trades daily with front‑office bursts near 10,000 tps, EOD time falling from 8h to 5h after introducing delta storage and dependency‑driven incremental updates, and using Python for orchestration with hot C++ compute building blocks for heavy math. Architecture is geo‑distributed with regional synchronous replicas, compute building blocks (lambda‑style) for parallel P&L, immutable audit streams for compliance, and a Kafka‑style event pipeline; the redesign proves resilient and performant while avoiding previous regulatory failures.
AI-written summary · about
chapters (10)
- 0:00Introduction and Overview
- 1:15Understanding Financial Trading Platforms
- 1:55Core Financial Entities and Their Importance
- 3:10Challenges in Scaling Financial Systems
- 6:46The Monolithic Nightmare
- 15:23Innovative Solutions for Scaling
- 24:40Compliance and Regulatory Requirements
- 26:38Why Python for Financial Systems?
- 27:32Key Principles and Future Outlook
- 30:11Conclusion and Final Thoughts
related talks
40mScaling Python - Bank Edition
51mClosing Keynote
45mPython at Massive Scale
37mAsyncio, websockets, and BTC sitting in a tree
01h 18mKeynote
40mBuilding highload ML powered service
16mEmpowering Innovation with the Julia Language
31mPython Doesn’t Have to Be Slow: Speeding Up a Large-Scale Optimization Algorithm