top of page

QuantLabsNet.com Public Quant Analytics Group

Public·929 members

Sharing My Experience: C++ Monolith vs. Modular Architecture - What Worked for Me


Hey folks,

Long-time lurker, first-time poster here. Want to share my recent experience building trading infrastructure because I could really use some community input.

Background:


 I've been running a Python-based trading system for market analysis and execution. Worked fine, but hit a wall with Rithmic/IBKR—they only allow one connection at a time, which became a bottleneck when trying to run multiple strategies.

My "solution":


 Built a unified C++ application to handle everything in one place—market data ingestion, order management, analytics. Seemed like the right call for performance and control.

What actually happened:

  • C++ codebase grew quickly

  • Every time I added a feature, previous fixes broke

  • AI tools (Claude, Kimi K3) kept regenerating code that overwrote my work

  • Compute costs for AI-assisted C++ development are no joke

  • Spent $1,000+ on this experiment

Current setup (post-pivot):

  • Python for analytics (kept this, it's solid)

  • C++ only for the Rithmic gateway (minimal footprint)

  • Redis as the message bus between components

My questions:

  1. Has anyone else experienced AI tools struggling with large C++ projects?

  2. How do you decide when to split services vs. consolidate?

  3. For those running trading systems—what's your approach to handling single-connection limitations?

Looking for genuine feedback. Is modular architecture the right call here, or am I missing something?

Thanks in advance.



13 Views
bottom of page