Machine Learning Engineer

Multiverse Computing; Barcelona; Hybrid; EUR 70,000 - 95,000; Full time

Benefits offered by this job

Private health insurance

Hybrid opportunity

Language classes

Relocation package

Visa sponsorship

Job summary

Multiverse Computing, a fast-growing deep-tech company in Barcelona, is hiring researchers and engineers to build scalable ML training stacks, data pipelines, and evaluation environments. You will work with transformer models, multi-GPU training, and deployment across NVIDIA stacks, while contributing to reproducible, containerized workflows.

You’ll collaborate with researchers and engineers in a hybrid setup, benefiting from a high-performance environment and opportunities to learn and teach

Qualifications

• 2+ years of industry experience building ML systems, including transformer models or LLMs.
• Strong Python software engineering: clean, maintainable, tested code; git, code review, CI.
• Strong PyTorch experience training or fine-tuning transformer models.
• Experience with multi-GPU and distributed training concepts (data, tensor, pipeline parallelism).
• Proficiency with Linux, containers and shared GPU clusters (Slurm).
• Fluent English; Spanish proficiency is a plus.

Responsibilities

• Build and maintain the training stack used by researchers (multi-node training, post-training pipelines).
• Create scalable data pipelines for curation, filtering, and synthetic data generation.
• Develop evaluation pipelines to ensure reproducibility across models and checkpoints.
• Build RL environments, verifiers, and reward pipelines for reasoning and tasks.
• Profile and optimize training/inference, GPU utilisation and memory.
• Package, optimize, and benchmark models for deployment on NVIDIA and edge hardware.
• Translate research ideas into reusable code and run experiments with the team.
• Maintain reproducibility through containers, configs, and ML experiment tracking.

Skills

Python

PyTorch

Transformers

Distributed training

Linux

Docker

Kubernetes

Megatron-LM

NeMo RL

English

Spanish

Education

Master's in Computer Science or related field

Tools

Docker

Kubernetes

Slurm

Megatron-LM

NeMo

Job description

Описание

Multiverse Computing Multiverse Computing is a fast-growing deep-tech company founded in 2019 and recognized by CB Insights as one of the 100 most promising AI companies globally. We are the largest quantum software company in the EU, with 300+ employees worldwide and growing, delivering advanced solutions trusted by leading global enterprises across several critical industries, including finance, energy, manufacturing, telecom, and industrial sectors. Our mission is to enable organizations to gain a meaningful competitive edge through cutting-edge AI and quantum technologies. You’ll work alongside world-leading experts, developing solutions that deliver real-world value for global clients. We are committed to fostering an inclusive, ethics-driven culture grounded in sustainability, diversity, and collaboration - a place where ambitious, curious people take ownership, push boundaries, and grow. Come and join us.

Задачи

• Build and maintain the training stack our researchers use every day: multi-node training and post-training pipelines (SFT, distillation, RL) built on Megatron-LM, Megatron Bridge and NeMo RL.
• Build scalable data pipelines for curation, filtering, deduplication, tokenization and synthetic data generation.
• Build and maintain evaluation pipelines, so benchmark results are reproducible and comparable across models and checkpoints.
• Build RL environments, verifiers and reward pipelines for reasoning, coding and agentic tasks, and keep rollouts fast and stable.
• Profile and optimise training and inference: GPU utilisation, memory, parallelism configuration, checkpointing, I/O.
• Package, optimise and benchmark models for deployment, including on non-NVIDIA and edge hardware.
• Turn research ideas (new architectures, compression methods, training recipes) into clean, tested, reusable code, and run experiments together with the research team.
• Keep our work reproducible through containers, configs, experiment tracking, CI, code review and documentation.

Требования

• Master’s in Computer Science, AI, Mathematics, Physics, Engineering or a related field, or equivalent industry experience.
• 2+ years of industry experience building ML systems, including at least 1 year hands-on with transformer models or LLMs.
• Strong Python software engineering: clean, maintainable, tested code; git, code review, unit and integration tests, CI.
• Strong PyTorch and hands-on experience training or fine-tuning transformer models.
• Experience running multi-GPU jobs and a working understanding of distributed training (data, tensor and pipeline parallelism; FSDP or DeepSpeed).
• Comfortable with Linux, containers (Docker or similar) and shared GPU clusters (Slurm or Kubernetes).
• Experience building data pipelines that process large datasets.
• Solid ML fundamentals: you understand what happens inside training, not just how to launch it.
• Good debugging and communication skills.
• Fluent English.
• We don't expect anyone to have all of these. Our research covers many areas and nobody on the team is an expert in all of them. If you are strong in some and new to others, we still want to hear from you.
• Experience with the NVIDIA stack: Megatron-LM, Megatron Bridge, NeMo RL, NeMo Gym, NeMo Curator.
• Hands-on post-training: SFT, DPO, or RL (GRPO, PPO), or frameworks such as verl, TRL or OpenRLHF.
• Inference engines (vLLM, SGLang, TensorRT-LLM) for serving, synthetic data generation or RL rollouts.
• Deployment on edge or alternative hardware: llama.cpp, ExecuTorch, ONNX Runtime, OpenVINO, vendor NPU SDKs.
• Model compression: quantization (GPTQ, AWQ, FP8), pruning, distillation.
• GPU performance work: profiling (Nsight, PyTorch profiler), Triton or CUDA kernels.
• Evaluation tooling: lm-evaluation-harness, lighteval, NeMo Evaluator.
• Experiment tracking and MLOps (MLflow, Weights & Biases), cloud platforms (ideally AWS).
• Experience with Physical AI tooling: robotics simulation (Isaac Sim/Lab, MuJoCo) or vision-language-action models.
• Spanish.
• Merged contributions to major open-source ML projects (PyTorch, Megatron-LM, vLLM, Hugging Face, etc.).
• Knowledge of tensor networks or quantum-inspired methods.
• Publications or patents.

Условия

• Indefinite contract.
• Equal pay guaranteed.
• Variable performance bonus.
• Signing bonus.
• We offer work visa sponsorship (if applicable).
• Relocation package (if applicable).
• Private health insurance.
• Eligibility for educational budget according to internal policy.
• Hybrid opportunity.
• Flexible working hours.
• Language classes and discounted lunch options.
• A high-performance, collaborative environment, operating at pace on cutting-edge technologies.
• Career plan.
• Opportunity to learn and teach.

As an equal opportunity employer, Multiverse Computing is committed to building an inclusive workplace.

The company welcomes people from all different backgrounds, including age, citizenship, ethnic and racial origins, gender identities, individuals with disabilities, marital status, religions and ideologies, and sexual orientations to apply.