The role you’ll play
You will work on improving both the quality and efficiency of our training pipelines as well as the inference performance of our Transformer based machine translation models.
The team you’ll be a part of
You will collaborate closely with Machine Learning Engineers, Data Engineers and Full stack Engineers, reporting to the team lead.
What you’ll be doing
- Optimizing our training pipelines and infrastructure
- Improving the relevance and quality of our training data
- Adopting new ML research into our architecture
- Optimizing our training data, to achieve better models with less data
- Ensuring our MT APIs scale and handle both real-time and batch workloads

In one year, you’ll know you were successful if
- If our machine translation models train on 2x the number of GPUs in ~0.5 the time
- If our inference speed is more than 50% faster on the same compute
- Our machine translation offers more language pairs and higher quality

Desired experience and competencies
What you’ll need to bring




- Master's degree in computer science and 3+ years of relevant work experience
- Expert knowledge of Python & PyTorch
- Experience training large neural networks on millions of examples on multiple GPUs
- Experience running neural networks in production
- Experience with Kubernetes (we use GKE)

This will make you stand out
- Experience with optimization of multi-GPU training
- Experience with training on huge NLP datasets
- Experience with building cost efficient E2E training pipelines
- Experience with inference optimization (e.g. Torchscript/TensorRT/Quantization)
- Experience with CI/CD pipelines and Terraform/Helm

About LanguageWire

Most important things to know about us:
- We eliminate language barriers so people understand each other. Everywhere.
- We are a tech company
- We are an AI company
- We are stable, and we are growing

More detail about us:
At LanguageWire, we are leaders in the world of Language Service Providers, but we are cooler than our c

📌 Machine Learning Engineer (España)
🏢 LanguageWire
📍 España