A big welcome and thank you to Capital One for joining the PSF as a Principal sponsor!
Capital One is also a PyCon 2019 Principal sponsor and is excited to share a few things with attendees, including a deeper look at their intelligent virtual assistant, Eno. Eno’s NLP models were built in-house with Python. Eno is a key component of the customer experience at Capital One, proactively looking out for customers and their money. Eno notifies customers about unusual transactions or duplicate charges, helping to spot fraud in its tracks. It also sends bill reminders and makes paying your bill as easy as sending a text or emoji; plus, its new virtual card capabilities let customers shop online without using their real credit card number.
The benefits they’ve seen by developing Eno with Python are numerous: fast time to market, the ability to prototype and iterate quickly, ease of integration with machine learning frameworks, and extensive support for everything we need (like Kafka and Redis). Plus, they see faster performance using Python's Asynchronous I/O.
For Capital One, sponsoring important industry conferences like PyCon brings a lot of benefits, like recruiting and brand awareness, but they’re here first and foremost for the community. By sponsoring PyCon, they feel they’re helping support, strengthen, and engage with the Python community.
Capital One sees the future of banking as real-time, data-driven, and enabled by machine learning and data science -- and Python plays a big role in that. They have embedded machine learning across the entire enterprise, from call center operations to back-office processes, fraud, internal operations, the customer experience, and much more. To them, machine learning not only creates efficiency and scale on a level not possible before, but it also helps give their customers greater protection, security, confidence, and control of their finances.
Python has been and will continue to be critical to advances in machine learning and data science, so they see a lot of exciting innovation, growth, and potential for the Python community. They hope to share back with the community some of their own insights, best practices, and broader work with Python.
As an open source first organization, Capital One has been working in the open source space for several years -- consuming and contributing code, as well as releasing their own projects. One example of an open source project they’ll be showcasing at PyCon is Cloud Custodian. Cloud Custodian is a tool built with Python to allow users to easily define rules to enable a well-managed cloud infrastructure in the enterprise. It’s both secure and cost-optimized and consolidates many of the ad-hoc scripts organizations have into a lightweight and flexible tool, with unified metrics and reporting.
They also developed a Javascript project called Hygieia, a single, configurable dashboard that visualizes the health of an entire software delivery pipeline. All their open source projects are on GitHub and their Python projects can be found here.
According to the Python Software Foundation and JetBrains’ 2018 Python Developers Survey, using Python for machine learning grew 7 percentage points since 2017, which is incredible. Machine learning experienced faster growth than Web development, which has only increased by 2 percentage points when compared to the previous year. Capital One is increasingly focused on using machine learning across the enterprise. One recent Python-based project is work they’ve done in Explainable AI. Their team created a technique called Global Attribution Mapping (GAM), which is capable of explaining neural network predictions across subpopulations. This approach surfaces subpopulations with their most representative explanations, allowing them to inspect global model behavior and essentially make it easier to generate global explanations based on local attributions. You can learn more about the open source tool they developed for GAM along with a recent whitepaper with more details.
Be sure to stop by their booth, #303, and get even more details about how they’re using Python.
Comments