Machine Learning Engineer

ML Performance Engineer – Low Latency Inference (Trading Environment)

Location : Chicago / NYC

Experience : 3-12 years (exceptional early-career profiles considered)

If you’re working on ML today, here’s the real question:

Are you optimising models for leaderboard metrics…

Or for speed under pressure?

This role sits inside a trading business where inference latency is measured in nanoseconds. Models don’t just need to be accurate; they need to fire first.

Firms compete for a few nanoseconds of edge over the market, which can lead to millions in profit.

What You’ll Do

• Optimise GPU-based inference pipelines for real-time decision systems
• Profile and eliminate bottlenecks across CPU, GPU, and network boundaries
• Tune memory layout, batching, and concurrency to reduce end-to-end latency
• Work directly with researchers and traders to productionise models
• Own performance in live environments where speed directly impacts outcomes

This is not research-only ML.

It’s ML under strict performance constraints.

You’ll Fit If

• Strong in Python and C++
• Hands-on with CUDA / GPU optimisation
• Comfortable profiling at kernel and system level
• Care about determinism, throughput, and hardware-software interaction
• Prefer performance problems over product roadmap debates
• Bachelors degree in a STEM field.

Finance experience isn’t required.

What matters is whether shaving off microseconds sounds more interesting than shipping another feature at a big lab.

If it does, it's worth a conversation.