From One Strategy to Many — and Still Building

Yangfeng joined Midas five years ago with a background in Computer Science from Soochow University. Today, as a Quantitative Developer, he leads projects and tackles complex engineering challenges across our trading systems.

What made you join Midas in the first place?

I was drawn to HFT because the work felt more meaningful and end-to-end. Rather than owning one small piece of a much larger system, I could work on problems from start to finish and see the impact of what I built.

The interview process also left a strong impression on me. The office felt international, and there was a real sense of energy among the people around me.

How has your role evolved since day one?

I now lead projects, make technical decisions, and take ownership of broader problems. The way we work has evolved significantly too.

LLMs can now generate large amounts of code from a short prompt, which makes defining the right constraints and understanding the problem clearly even more important.

Latency, meanwhile, remains critical. We continuously optimize it, but knowing what to optimize requires precise measurement — measure first, then optimize.

Looking back, what’s changed the most over the past five years?

We now trade across more markets and products, and our technology has evolved along with us.

We automate well-defined processes — from notifications to testing and deployment — so we can spend less time on repetitive work and more time on harder problems.

We also keep exploring new hardware, from CPUs to NICs. At one point, we achieved latency comparable to Solarflare-class network cards with a much lower-cost setup.

After five years, what keeps you here?

The work has stayed end-to-end.

We’ve grown significantly, but that growth hasn’t fragmented the work. Instead, we automate the well-defined parts so that important problems remain ownable from one end to the other.

The market is reflexive, the game is recursive, and the bottlenecks keep moving. Five years in, there are still new problems to solve — and still more to build.

What advice would you give someone joining Midas today?

Three things: low ego, honesty, and attention to detail.

Low ego — the market ultimately decides who is right. It means listening to your colleagues rather than defending your own ideas, and being willing to let go of a hypothesis even after spending a long time validating it.

Honesty — bad news should be shared, and bugs should be reported as soon as they’re found. That’s expected and accepted here.

Attention to detail — it helps us identify issues quickly, find solutions faster, and evaluate results more accurately.