Context
This system was built for a trading environment where milliseconds had direct financial impact. The goal was to detect NFT arbitrage opportunities on Solana and execute before the opportunity disappeared.
Main challenge
The problem was not just writing a trading strategy. The system had to react faster than competitors and remain stable under real market conditions.
What I worked on
- Bot architecture for identifying and executing arbitrage opportunities
- Global server deployment to reduce round-trip time from multiple regions
- Redis-based in-memory storage for hot path data
- Linux performance tuning and lower-level optimization work
- Customization of Solana libraries to remove unnecessary overhead
Outcome
The system generated $300K in net profit across roughly $12M in trading volume.
It also reached sub-100ms reaction times, which required consistent performance across infrastructure, networking, and application code rather than one isolated optimization.
What made this work valuable
This project is a good example of how I approach performance-sensitive systems. I like the combination of measurement, iteration, and practical engineering needed to make a system fast in the real world instead of only in theory.
Engineering decisions
The main lesson from this project is that latency work is rarely about one optimization. The meaningful gains came from combining infrastructure placement, in-memory data handling, system tuning, and library-level changes into one coherent path.