AI Engineer - Mid / Senior

Specialization: Model Optimization and Hardware Acceleration

Job Description

Our partner is looking for experienced AI Engineers with strong expertise in neural network internals, model optimization and hardware execution.

The ideal candidate will have practical experience working directly with neural network architectures and their implementations, improving inference performance, memory efficiency and computational resource utilization.

This position is particularly suited to engineers who enjoy working beyond traditional AI application development, with a focus on understanding how neural networks execute, identifying performance bottlenecks and modifying model implementations to improve efficiency on specialized hardware.

You will work closely with AI software, compiler and hardware engineering teams to optimize neural networks for advanced processor platforms.

Location: Barcelona, Spain
Working model: Hybrid

Responsibilities

• Research, implement and optimize neural network architectures, including transformers, CNNs, vision models and RNNs.

• Work directly with neural network internals, modifying model components, operations and execution behavior to improve performance.

• Implement optimization techniques such as quantization, pruning, compression and reduced precision computation.

• Analyze and optimize inference latency, throughput, memory consumption and computational efficiency.

• Investigate performance bottlenecks using benchmarking, profiling and debugging tools.

• Evaluate the relationship between model accuracy, computational cost and hardware resource consumption.

• Adapt and optimize models for execution on CPUs, GPUs and specialized AI accelerators.

• Develop and maintain C++ and Python components supporting AI model optimization and execution.

• Work with PyTorch and relevant inference frameworks to investigate model behavior and improve performance.

• Collaborate with software, compiler and hardware engineering teams to improve AI workload execution.

• Validate optimization results through reproducible performance measurements and accuracy testing.

• Support integration and deployment of optimized AI models into production environments.

Mandatory Requirements

• Practical experience modifying neural network internals or model execution, beyond training, adapting and deploying AI applications.

• Strong understanding of transformers, CNNs, vision models and RNNs, including their internal operations and computational characteristics.

• Professional C++ programming experience, preferably involving performance optimization, numerical computing, machine learning or systems software.

• Strong practical experience with PyTorch, including model implementation and architecture modification.

• Experience implementing neural network optimization techniques, such as quantization, pruning or model compression.

• Ability to identify and investigate performance bottlenecks affecting model inference.

• Experience analyzing inference latency, throughput, memory consumption or computational efficiency.

• Understanding of how model architecture and implementation affect execution on different hardware platforms.

• Experience using Git and working with CI/CD processes.

• Strong technical problem solving and debugging skills.

• English proficiency at C1 level.

Important: We are particularly interested in candidates who can demonstrate direct responsibility for optimizing neural network implementations, explain the technical changes they made and describe how performance improvements were evaluated.

Optional / Desired

• RISC-V architecture or processor experience

• Experience with AI accelerators and embedded inference

• MLIR, LLVM or compiler optimization

• ONNX Runtime and execution providers

• CUDA, ROCm or Triton

• Custom kernel development

• TensorRT and TensorRT LLM

• OpenXLA, TVM or IREE

• vLLM, SGLang or DeepSpeed

• PyTorch internals, operators or execution graphs

• Mixed precision computation, operator fusion or memory optimization

• Profiling tools such as Perfetto, NVIDIA Nsight or equivalent

• TensorFlow

• pybind11

• GDB and software debugging

• Experience measuring accuracy and performance tradeoffs

What’s in it for you?

• Flexible working hours

• Hybrid model, 4 days office, 1 day from home

• 1 week per year working from anywhere

• 25 days annual leave plus December 24 and 31

• Monthly food allowance

• Private medical insurance

• Relocation bonus

• Housing search support

• Visa support if required

• Family relocation support

• Virtual shares

• Language classes

• Tax incentive, subject to eligibility