Machine Learning Performance Engineer – Training
An engineering role speeding up large-scale model training for quant researchers at a trading firm. It spans distributed training, GPU kernels and cluster efficiency.
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.
- Benchmark training workloads across accelerators
- Tune distributed training and multi-GPU communication
- Improve data loading, checkpointing and utilisation
- Develop GPU kernels for research models
- Work with HPC teams on scheduling and observability
What they ask for
Required
- 3+ years optimising ML training at scale
- PyTorch or JAX internals
- Python and C++
- GPU kernel work (CUDA, Triton)
- Distributed training (NCCL, FSDP, DeepSpeed)
- Profiling tools
- Cluster networking basics
Preferred
- Transformer and time-series training
- Kubernetes, Slurm or Ray
- Fault-tolerant training
- Custom accelerators
- Education
- Not stated.
- Experience
- 3+ years optimising ML training workloads; finance not required
- Languages and tools
How to prepare
- Train a model with FSDP or DeepSpeed and analyse scaling
- Profile a training run with Nsight or PyTorch Profiler
- Write a custom CUDA or Triton kernel
- Learn NCCL collective operations
What to study
- Financial Machine Learning
- Modern C++ for Quantitative Finance
- Quantitative Data Engineering
- Trading Systems and Low-Latency Infrastructure
The NLP and ML Quant Researcher roadmap sets out the stages for this kind of role, with free resources for each.
Open the roadmapApply
Read the full posting on Tower Research Capital’s site before applying: requirements, location and pay can change after we check them.
View the posting on Tower Research Capital’s site ↗
SelfMadeQuant wrote this summary from the employer’s public posting, last checked 6 November 2026. SelfMadeQuant isn’t affiliated with Tower Research Capital and doesn’t handle applications.