ML Engineer 4: Production AI at Scale

Capital One; San Jose (CA); On-site; USD 215,000 - 246,000; Full time

Job summary

Capital One seeks a Machine Learning Engineer 4 for the MLX team in San Jose, CA. You will design, build, and deploy ML models that solve real business problems, collaborating with Product and Data Science teams to deliver scalable AI solutions.

You will work with PyTorch or TensorFlow, Spark, and Kubernetes, maintain production models, build data pipelines, and implement CI/CD practices while upholding Responsible AI standards.

Qualifications

• Bachelor's degree in Computer Science, Machine Learning or related quantitative field.
• At least 4 years of programming in Python/Java/Golang/C++.
• 4+ years of ML experience with PyTorch or TensorFlow and data libraries.
• 4+ years operating large-scale distributed systems (Spark, Ray).
• 2+ years deploying ML in production using cloud platforms and Kubernetes.

Responsibilities

• Design, build, and deliver ML models and components with Product and Data Science teams.
• Inform ML infrastructure decisions including data, features, and model selection.
• Collaborate in an Agile team to enable big data and ML applications.
• Retrain, monitor, and maintain models in production.
• Leverage cloud architectures to deliver ML models at scale.
• Construct optimized data pipelines for training and evaluation.

Skills

Python

Java

Golang

C++

Education

Bachelor's degree in CS/ML or related field

Master's or Doctoral degree preferred

Tools

PyTorch

TensorFlow

Pandas

NumPy

Scikit-learn

Spark

Ray

Kubernetes

Job description

Capital One seeks a Machine Learning Engineer 4 for the MLX team in San Jose, CA. You will design, build, and deploy ML models that solve real business problems, collaborating with Product and Data Science teams to deliver scalable AI solutions.

You will work with PyTorch or TensorFlow, Spark, and Kubernetes, maintain production models, build data pipelines, and implement CI/CD practices while upholding Responsible AI standards.