Compiler Engineer

Oho Group; Toronto; On-site; CAD 110,000 - 170,000; Full time

Job summary

Oho Group in Toronto is seeking a Compiler Optimization Engineer to shape the part of the compiler that turns model structure into efficient execution. You’ll work between model import and target code generation, making decisions about representations, transformations and preparation for execution.

You’ll collaborate with frontend and backend engineers, leverage hardware measurements to spot optimization opportunities, and help evolve the IR/pipeline as new model patterns emerge.

Qualifications

• Strong C++ and hands-on experience with compiler transformations or IRs.
• Solid understanding of dataflow, optimization legality and performance tradeoffs.
• Ability to connect graph and layout decisions with memory behavior on accelerators.
• Experience with MLIR, XLA, affine or polyhedral analysis or ML framework internals is a plus.

Responsibilities

• Developing graph transformations that reduce redundant work and data movement.
• Improving fusion decisions and how tensor layouts propagate through a program.
• Evolving the IR and pass pipeline as new model patterns emerge.
• Using hardware measurements to identify missed optimization opportunities.
• Collaborating with frontend and backend engineers to keep compiler interfaces clear.
• Building validation that protects correctness as optimisations become more ambitious.

Skills

C++

Compiler transformations

IR design

Dataflow & optimization

Memory behavior on accelerators

Tools

MLIR

XLA

Affine/Polyhedral analysis

Job description

I’m working with a venture-backed AI infrastructure startup building a compiler and runtime platform designed to improve accelerator efficiency and give developers greater flexibility over where their workloads run.

They’re looking for a Compiler Optimization Engineer to shape the part of the compiler that turns model structure into better execution.

The opportunity

You’ll work between model import and target code generation, making decisions about how computations are represented, transformed and prepared for execution.

Your work will include:

• Developing graph transformations that reduce redundant work and unnecessary data movement.
• Improving fusion decisions and how tensor layouts propagate through a program.
• Evolving the IR and pass pipeline as new model patterns emerge.
• Using hardware measurements to identify missed optimisation opportunities.
• Working with frontend and backend engineers to keep compiler interfaces clear.
• Building validation that protects correctness as optimisations become more ambitious.

What would make you a strong fit

• Strong C++ and practical experience developing compiler transformations or IRs.
• A clear understanding of dataflow, optimisation legality and performance tradeoffs.
• The ability to connect graph and layout decisions with memory behaviour on accelerators.

Experience with MLIR, XLA, affine or polyhedral analysis, or ML framework internals would be particularly relevant.

The appeal is substantial technical ownership: your decisions will influence how models are compiled and how effectively they use different hardware. You’ll join an experienced compiler and systems team, with equity participation and room to develop new approaches.