In the previous article, we discussed how we evaluated Cloud Bigtable for Sift, how we prepared our systems for the zero downtime migration from HBase, and the overall data migration strategy. In this post, we present the unknown unknowns...
As the leader in Digital Trust & Safety and a pioneer in using machine learning to fight fraud, we regularly deploy new machine learning models into production. Our customers use the scores generated by our machine learning models to...
In the previous article, we described how we assessed the fitness of Bigtable to Sift’s use cases. The proof of concept (POC) also helped us to identify some parts of our system that needed to be modified for the...
In the previous article, we presented the challenges that prompted us to migrate away from HBase on Amazon Web Services (AWS). In this article, we describe how we went about evaluating Cloud Bigtable as a potential datastore for Sift....
At Sift, we have built a scalable and highly available technology platform to enhance trust and safety in the digital world. We ingest, process, and store petabytes of data from our customers from around the globe. When our customers...
We are excited to announce the public release of a package we’ve been using internally at Sift to handle our client-side data fetching for over two years now: resourcerer. Resourcerer contains a React higher-order component (HOC) called withResources that...
AWS notified Sift on September 18th, 2018, of an issue where AWS incorrectly labeled Amazon Machine Images associated with Sift’s account. AWS incorrectly labeled AMIs as encrypted when the images were unencrypted. Sift engineers investigated and confirmed that the...
Engineering Management Series The past few years have been a time of incredible growth at Sift and there is no sign that we are slowing down. To continue delighting our ever growing list of customers, we’ll be scaling our...
A few years ago, the advice for keeping your page fast was to keep your page weight down. These days, however, we know that it’s not necessarily how big your page is, it’s how, when, and if you send...
The rapidly evolving device and browser landscape allows us to collect increasingly rich data via our snippet. Because we host the snippet and our customers fetch it from us dynamically, expanding its functionality requires a strict eye on compatibility...
In the adversarial and constantly changing world of fraud detection and trust enablement, keeping up with the most recent online patterns is crucial to success. Sift Science uses machine learning (ML) models to rate the riskiness of online events...
In this blog post we detail how Sift has begun leveraging deep learning (in the form of RNNs) to improve our ability to detect fraud.
At Sift Science, engineers train large machine learning models for thousands of customers. We need processes and tools to do this consistently and reliably. In this blog, we discuss how we use Apache Airflow to manage Sift’s scheduled model training...
By now, you’re probably familiar with the benefits of splitting up your app’s javascript into several smaller files through code splitting as opposed to serving a single behemoth. These days, code splitting is no longer even just a Webpack...
TL;DR: We can transform the score distributions of new models to match those of old models, while preserving the new model’s ROC curve. This allows model changes to improve accuracy without disrupting automated systems triggered by fixed score thresholds....
Sift Science’s iOS SDK enables mobile applications to send us device properties and application lifecycle events for use in fraud detection. There are special considerations to be taken in designing data-intensive tracking SDKs for mobile devices with limited bandwidth,...
This summer we launched Workflows, a new product that allows fraud managers to build and update their fraud processes without needing to write any code. Along with Decisions, which lower the bar for implementing automatic fraud prevention, and Review...
In the first two posts of this three-part series we discussed how we ensure correctness when conducting and analyzing machine learning experiments at Sift. Determining how to construct experiments and evaluate their results correctly is great, but it is...
In the first post on ML Experiments at Sift Science, we described how we minimize bias in our offline experimentation framework to ensure we run representative experiments. Once this is achieved, the next step is to determine how to...
At Sift Science we use machine learning to prevent various forms of abuse on the internet. To do this, we start with some data provided by our customers: Page view data sent via our Javascript snippet Event data for...
I remember being intrigued by Smashing Magazine’s post this summer about using PJAX to effectively turn static markup websites into single-page apps (SPAs). Turbolinks even turned the idea into a library! But I was especially interested in the section...
In the first post in this series, we looked at how coding within the React paradigm could lead to poor browser performance. We fixed a couple of those issues in the context of an auto-height, slide-down animation, but let’s...
We’re happy to announce that we’re open sourcing one component of our Kafka operations toolkit: an algorithm that computes a new partition assignment to brokers while obeying the following properties: Data movement is minimized Partitions for each topic are...
This post is the first in a series that describes our experience in adopting Apache Kafka. Here at Sift Science, we have introduced Kafka as the messaging layer between our distributed services. In particular, we treat Kafka topics as...
Adventures in React Performance DebuggingRecently I read Benchling’s 2-part series in debugging performance issues in React, and it really echoed the issues and solutions that I’ve been working through on the Sift Science Console. So I was inspired to chime...
At Sift Science, we use a variety of popular machine learning models to detect fraud for our customers. However, until recently we relied exclusively on a combination of linear models and sophisticated feature engineering. As we were reaching the...
Here at Sift Science, we just completed another big step in our ongoing marketing site redesign, overhauling the homepage and replacing old landing pages with [prettier, responsive, and more performant ones][1]. While the big performance improvements aren't quite ready...
We're adding random decision forests to our machine learning solution, so get ready for an 18% improvement in Sift Score accuracy!This week, we launched an entirely new machine learning model called random decision forests, which will work alongside...
In the first post of this series, we gave an overview of Sift Science’s architectural migration to React and Dropwizard. We followed up with some best practices for scaling React in a production setting and...
On May 7th, I presented at HBaseCon, demonstrating how Sift Science leverages HBase and its ecosystem in powering our machine learning infrastructure. In case you missed the talk, I’ll lay out the main points here.There are three...
A little less than a year ago, the Sift Science Console Team decided to migrate its Backbone and Marionette Views to ReactJS see also our post on specific React tips we learned along the way....
Sift Science has been using React in production for almost a year now. In that time, we grew our application from a Backbone + React frankenstein hybrid app into one fairly large hierarchy of React components. In this post,...
Two years ago, we publicly launched our first fraud product with the goal of making it easy for anybody to leverage the same machine learning technology that protects the largest internet retailers. That product had a very...
Last Thursday, Sift Science was gracious enough to host API Craft San Francisco. API Craft SF is a meetup group I organize that focuses on bringing together API practitioners that care about their craft: architects, designers, developers, testers,...