Exploring Machine Learning at Capital One.

Welcome to Unrival.AI's 'Technology Stack' Series, powered by our Unrival GPT Model - Zapp.

Zapp Icon - UnrivalGPT

Meet Zapp

Zapp is a AI model built by the team of Machine Learning and Deep Learning Engineers at Unrival.AI. Zapp has extensive knowledge on the technologies used by Fortune 500 companies and what they use these technologies for. In this article, Zapp shares its knowledge on accounts via this blog.

Hey Zapp, can you tell us about Machine Learning at Capital One?
Zapp Icon
Machine Learning (ML) at Capital One.

Machine Learning at Capital One is a central focus area that is seen as a significant opportunity to transform the financial services industry. The organization's commitment to Machine Learning is supported at the highest levels, including the CEO, the Board of Directors, and the executive committee.

Here are some key aspects of Machine Learning at Capital One:

// Strategic Importance //
Machine Learning is leveraged to create innovative products and services, enhance customer experience, and improve operational efficiency across various lines of business within Capital One.

// Innovation and Research //
Capital One is actively engaged in exploring and implementing cutting-edge Machine Learning technologies, including explainable AI, generative AI, and large language models.

// Infrastructure and Platforms //
The company has invested in building robust ML platforms and infrastructure, such as an enterprise AI/ML Platform, an internal Feature Platform, and the Apollo team's Business Data Platform, to support the development and deployment of ML models at scale.

// Talent and Expertise //
Capital One employs a diverse team of Machine Learning Engineers, Data Scientists, and other specialists who work on a wide range of ML applications, from credit risk modeling and fraud detection to customer service enhancements and marketing optimization.

// Cloud Computing //
Following a complete migration to the cloud, Capital One leverages cloud computing capabilities to manage data at scale and enhance its ML efforts, primarily using AWS services.

// Community and Open Source //
The organization contributes to the ML community through conference presentations, papers, blog posts, open source contributions, and patents.

// Regulatory Compliance //
As a financial institution, Capital One builds systems and solutions within a highly regulated environment, ensuring that its ML practices comply with industry standards and regulations.

Overall, Machine Learning at Capital One is about harnessing the power of data, advanced analytics, and cloud technologies to drive innovation and deliver real-world solutions that benefit customers and the business.
What are tools and technologies being used at Capital One for Machine Learning?
Zapp Icon
MLOps Tools and Technologies at Capital One.

Capital One utilizes a robust stack of tools and technologies to support its machine learning initiatives, including:
- Programming Languages: Python, Scala, Java
- Machine Learning Libraries: PyTorch, Scikit-Learn, XGBoost
- Data Processing and Analytics: Apache Spark, Dask
- Workflow Orchestration: Apache Airflow
- Containerization: Docker, Kubernetes
- Cloud Services: AWS (including EC2, ECS, Fargate, Lambda)
- Data Storage: Snowflake, Postgres, Redis
- Streaming and Messaging: Kafka
- In-Memory Computing: Hazelcast
- Model Serving: Seldon
- Monitoring and Logging: Tools for real-time assessment of ML applications
- CI/CD: Jenkins pipelines, Artifactory
- Version Control: Git

These tools are part of a comprehensive ecosystem designed to support the development, deployment, and scaling of machine learning models and applications across the organization.
Can you expand more on how Capital One are using PyTorch?
Zapp Icon
PyTorch at Capital One.

At Capital One, PyTorch is utilized as a core tool within the machine learning and data science ecosystem for a variety of purposes:
Real-time Neural Collaborative Filtering: PyTorch has been deployed to power a real-time neural collaborative filtering recommendation engine. This engine is a critical component of Capital One’s mobile app experience, resembling a "newsfeed" where personalized content is delivered to users.
Deep Learning-Based Embedding Technology: The data science team, particularly those focused on digital personalization, leverage deep learning-based embedding technology to convert complex data into embedding vectors. These vectors are then used for multiple downstream use cases, such as recommendation systems and customer experience enhancements.
Research and Development: PyTorch's dynamic computation graph and extensive library of modules and tools make it suitable for research and development of new machine learning models, including experimentation with different architectures and rapid prototyping.
Model Training and Deployment: For training complex models that require high-performance computing, PyTorch provides the flexibility and efficiency needed. Once models are trained, they can be deployed into production environments where real-time inference is critical.
Collaboration with Other Tools: PyTorch is often used in conjunction with other technologies in Capital One's tech stack, such as AWS for cloud computing resources, and it may be integrated with other machine learning and data processing tools like Spark and Dask.The use of PyTorch at Capital One is indicative of the company's commitment to leveraging cutting-edge machine learning technologies to enhance customer experiences and improve operational efficiency.

Got more questions? Join our waitlist

Request early access