About this role
Engineers who enjoy looking past high-level abstractions to understand how code actually executes on custom silicon will find a natural fit here. Robots and Pencils operates as an AWS Partner, embedding directly with enterprise clients to build production-ready AI systems that move beyond experimental proofs of concept. Team members focus heavily on training and operating models on Amazon SageMaker utilizing AWS Trainium chips, working from the hardware and compiler level all the way up through PyTorch training pipelines.
Day-to-day work centers on optimizing training performance, managing device-specific compilation, and troubleshooting issues that emerge at the hardware layer. The position exists to help clients bridge the gap between custom silicon capabilities and reliable, cost-aware production workloads. Ideal candidates bring deep framework knowledge alongside the curiosity to reason about memory, throughput, and compilation behavior on specialized accelerator infrastructure.
What you will do
- Train and operate models on Amazon SageMaker utilizing AWS Trainium as the underlying compute infrastructure.
- Write and optimize PyTorch training code with a clear understanding of compiler behavior, NeuronCore architecture, and memory tradeoffs.
- Diagnose training run issues stemming specifically from hardware or compiler behavior rather than model logic alone.
- Translate high-level client requests into working, cost-aware training pipelines from end to end.
- Tune distributed training runs for throughput and cost on SageMaker training infrastructure.
- Collaborate directly with client and internal engineering teams to scope and deliver production training workloads.
What you need
- Bring strong, hands-on PyTorch experience, ideally including multi-device or distributed training.
- Demonstrate production experience with Amazon SageMaker for training or inference.
- Show comfort working close to the hardware layer with an understanding of device-specific compilation and accelerator debugging.
- Possess solid Python fundamentals and the ability to operate effectively in a client-facing engineering environment.
Nice to have
- AWS Trainium or Inferentia (Neuron SDK) experience
Skills and tech stack
Questions candidates ask
What is the work setup for this position?
The position is remote based in Canada. Team members work forward-deployed and embedded directly with client teams.
What core technologies will I be working with?
The stack centers on PyTorch training code, Amazon SageMaker, and AWS Trainium custom silicon. Engineers also utilize the Neuron SDK and Python.
What kind of projects does the team build?
The team builds production AI systems for enterprise clients. The focus is on delivering real engineering workloads rather than proofs of concept that remain in notebooks.
Do I need prior experience with AWS Trainium?
AWS Trainium or Inferentia experience is listed as a strong plus rather than a strict requirement. Candidates with deep PyTorch skills and a track record of learning new hardware targets quickly are encouraged to apply.
The application is on the official Robots and Pencils careers page.
Apply on company site