Quant jobs / Susquehanna International Group (SIG)

Machine Learning Engineer – Inference Optimisation

An engineering role making production machine-learning models run fast on GPUs, mainly transformer-style models. It supports quantitative researchers by turning their models into low-latency serving systems.

What you’d work on

Our reading of the posting: what a new hire in this role would most likely spend their first months on.

Profile model inference end to end and find bottlenecks in batching, memory movement and serving
Tune inference runtimes and settings to improve latency, throughput and GPU use
Build benchmarking tools to compare model variants and hardware
Advise researchers on model design choices that are cheap to deploy

What they ask for

Required

Production ML inference experience
Python plus a systems language such as C++, Rust or Go
PyTorch internals: export, tracing, compilation, profiling
Reasoning about latency, throughput and GPU memory
Trade-off judgement

Preferred

Quantisation, pruning and distillation
CUDA, Triton or similar GPU programming
Kubernetes or GPU cluster experience
Operator fusion and model compilation

Education: No degree level stated; technical background such as maths, physics, CS or engineering is preferred
Experience: Experienced hire with production ML inference work
Languages and tools: Python
C++
Rust
Go
Java
C#
PyTorch
CUDA

How to prepare

Take a small transformer, export it and profile it with PyTorch tools, then cut latency by batching and quantisation
Learn how GPU memory and kernel launches limit throughput
Write a short benchmark report comparing runtimes and settings
Study low-latency serving patterns such as dynamic batching

What to study

Apply

Read the full posting on Susquehanna International Group (SIG)’s site before applying: requirements, location and pay can change after we check them.

View the posting on Susquehanna International Group (SIG)’s site ↗

SelfMadeQuant wrote this summary from the employer’s public posting, last checked 6 November 2026. SelfMadeQuant isn’t affiliated with Susquehanna International Group (SIG) and doesn’t handle applications.