ML Compiler Engineer

Type: Contract (6-9 months)

Location: Barcelona, Spain

​

Role Overview:

We are seeking a Front-End Compiler Engineer to design, develop, and scale the compiler front-end for our AI/ML stack. This role focuses on building Python-based model conversion pipelines that translate models from popular ML frameworks such as ONNX, TensorFlow, and PyTorch into our internal Intermediate Representation (IR) .

The ideal candidate will work extensively on graph-level representations and optimizations , support modern deep learning architectures (including LLMs) , and build robust testing infrastructure to ensure correctness, performance, and long-term maintainability of the compiler front-end.

Key Responsibilities:

• Design, develop, and maintain Python-based front-end converter modules to ingest models from ONNX, TensorFlow, and PyTorch into an internal IR.

• Implement graph construction, transformation, and IR lowering pipelines as part of the compiler front-end.

• Analyze computation graphs and implement graph-level optimization passes , such as operator fusion, simplification, and canonicalization.

• Build and extend pattern-matching and graph-rewriting frameworks for scalable and maintainable optimizations.

• Work on model decomposition and conversion of key building blocks used in LLMs , including attention mechanisms, MLPs, normalization layers, and embeddings.

• Leverage and integrate tools from ONNX Runtime for model parsing, validation, and conversion workflows where applicable.

• Develop and maintain Python-based testing infrastructure for correctness validation, operator coverage, regression testing, and CI integration.

• Debug and resolve issues across model ingestion, conversion, graph optimization, and IR generation stages.

• Collaborate with backend compiler, runtime, and performance teams to ensure end-to-end model correctness and efficiency.

Required Skills & Experience:

• Strong Python programming skills (mandatory) with an emphasis on clean, modular, maintainable, and well-tested code.

• Solid understanding of compiler fundamentals , including:

- Intermediate Representations (IRs)

- Graph-based computation models

- Transformation and optimization passes

• Hands-on experience with ML frameworks , including ONNX, TensorFlow, PyTorch , and exposure to Caffe .

• Practical experience in graph parsing, transformation, and optimization for ML models.

• Familiarity with modern ML architectures, particularly CNNs and Transformer-based models .

• Experience building or contributing to testing frameworks for compilers, ML systems, or large Python codebases.

• Strong debugging and problem-solving skills across complex, multi-stage pipelines.

Good to Have:

• Familiarity with MLIR-based front-ends and dialects , such as:

- TOSA

- StableHLO

- Torch-MLIR

• Exposure to AI compiler stacks, hardware backends, or accelerator targeting.

• Experience working with large-scale models or production ML inference/training pipelines.