Alpha Vertex: Smarter investing with artificial intelligence

About Alpha Vertex

Alpha Vertex builds cognitive technologies to change the way finance professionals discover and analyze investment opportunities, combining artificial intelligence with human insight to identify hidden relationships in financial information and drive better investments.

Industries: Financial Services
Location: United States

Alpha Vertex provides advanced analytical capabilities to the financial community, using 64-core VMs on Google Cloud Platform with Intel Xeon “Skylake” processors, reducing training times by 20%.

Google Cloud Results

  • Improves prediction accuracy by introducing hundreds of new machine learning models and adding dozens more every month
  • Scales easily, enabling 7-person team to create and support powerful big data solutions
  • Reduces model training time by 20% with Intel Xeon “Skylake” processors
  • Supported 5x growth in compute needs without exceeding cost of previous provider

Training 20K machine learning models at once

In financial markets, “alpha” is a measure of investment performance, or the active return compared to the market over the same time period. Investors naturally try to forecast opportunities that maximize alpha, which involves analyzing and coordinating an ever-growing amount of information. While investment research has traditionally been a time- and resource-intensive process, artificial intelligence (AI) has the potential to change the way finance professionals discover investment opportunities.

Alpha Vertex is seizing this opportunity by providing analysis and forecasting for financial markets with help from AI. By building AI-powered technologies that work with people to predict financial outcomes and identify trends before they become obvious, Alpha Vertex helps clients create custom investment models that deliver the highest returns.

“Google Kubernetes Engine is like magic for us. It’s the best container environment there is. Without it, we couldn’t provide the advanced financial analytics we offer today. Scaling would be difficult and prohibitively expensive.”

Michael Bishop, CTO and Co-Founder, Alpha Vertex

Alpha Vertex has compiled a growing knowledge base of linked data that tracks thousands of global stocks, changing market conditions, and other financial and international news across North America, Europe, and Asia. Every day, it adds as many as 2 million more pieces of unstructured text data—government statistics, press releases, trade journals, and executive interviews, among other sources. It then runs machine learning models on the data to develop predictive analyses. As data grows and new models are added and trained, insights get deeper and more reliable.

Training 20,000 machine learning models concurrently requires substantial compute power and memory. To meet these requirements, Alpha Vertex built the company on bare-metal cloud servers. However, managing and scaling the environment was increasingly challenging, taking most of CTO and Co-Founder Michael Bishop’s time.

“I was working 90 hours a week just to make sure our infrastructure would work,” he says. “At the same time, we needed better performance for model training. We operate 24x6 and have a fixed window in which we can retrain models and deliver predictions, so every hour is critical.”

“Using Google Cloud Platform, our machine learning training times have improved by 20%, giving us one-fifth more compute power to redirect to experimentation and innovation. We built a Kubernetes cluster of 150 64-core Intel Xeon Skylake processors in just 15 minutes.”

Michael Bishop, CTO and Co-Founder, Alpha Vertex

To enhance its service, Alpha Vertex tried Google Cloud Platform, and was immediately impressed with the performance it could achieve using Google Compute Engine 64-core machine types with Intel Xeon “Skylake” processors. Then it discovered that it could use Google Kubernetes Engine to automate container management and orchestration, giving Alpha Vertex just-in-time compute power regardless of job size. Convinced of the advantages, the company moved the bulk of its infrastructure to Google Cloud Platform.

“Google Kubernetes Engine is like magic for us,” says Michael. “It’s the best container environment there is. No other back-end system has had such a dramatic impact on our success. Without it, we couldn’t provide the advanced financial analytics we offer today. Scaling would be difficult and prohibitively expensive.”

20% more capacity for innovation

Alpha Vertex initially saw a 15% improvement in training times for its machine learning models after moving to Google Cloud Platform. Using Intel optimized libraries for machine learning reduced training times even further. Both the processor itself and the improved memory bandwidth contribute to the performance boost under concurrent workloads.

“Using Google Cloud Platform, our machine learning training times have improved by 20%, giving us one-fifth more compute power to redirect to experimentation and innovation,” says Michael. “We built a Kubernetes cluster of 150 64-core Intel Skylake processors in just 15 minutes.”

By shrinking model training and inference time, Alpha Vertex has added several hundred new inputs to its machine learning models, resulting in improved accuracy of predictions. It continues to add dozens of new models every month. As additional libraries and programs are refactored to take full advantage of the “Skylake” processor, Alpha Vertex expects to see further performance improvements. Memory-bound workloads benefit as well, with nearly twice the available memory bandwidth of the previous generation Intel Xeon.

“We’ve seen significant performance improvements in database loads and when running large Apache Spark jobs on Intel Xeon Skylake,” says Michael. “We get faster boot times with Google Cloud Platform than we’ve seen on other providers, especially when bursting up Kubernetes clusters. We also found the Google network to be the highest performing under demanding conditions.”

Saving time and costs

In addition to Google Cloud Platform compute resources, Alpha Vertex uses other managed cloud services: Google Cloud Storage for its data lake and long-term storage, Google BigQuery for data warehousing, Google Cloud Dataprep for data preparation and cleansing, Google Cloud Dataproc for data processing, and Google Cloud Pub/Sub for real-time messaging between applications. The entire cloud infrastructure is monitored with Google Stackdriver for reporting and metrics.

“We’re a small business that needs big infrastructure resources to make our services possible,” says Mutisya Ndunda, CEO and Co-Founder of Alpha Vertex. “Relying on an ecosystem of cloud services from Google Cloud Platform allows us to keep all our team members focused on developing and improving our core product.”

As a result, Michael no longer needs to babysit servers, and instead can devote his time to more strategic tasks. “It’s a night and day difference,” he says. “I wouldn’t want to go back.”

To reduce costs, Alpha Vertex uses Preemptible Virtual Machines—highly affordable, short-lived VMs—for most of its Google Cloud Platform workloads. If a VM gets preempted, Google Kubernetes Engine simply reschedules the workload onto a new node created by the auto-scaler.

“Given Google’s scale, we have yet to experience a situation in which preemptible nodes were unavailable,” says Michael. “Using Preemptible VMs with per-second billing contributes to a fantastic bottom line cost for us on Google Cloud Platform.”

“By choosing Google Cloud Platform, we can remain a small business and still be global and cutting edge. As Google continues to provide us with early access to best-in-class technology such as Intel Xeon Skylake processors to run live alpha testing at scale, we can innovate even faster.”

Mutisya Ndunda, CEO and Co-Founder, Alpha Vertex

Collaboration built for cloud

To help drive innovation, Alpha Vertex migrated from a cloud version of legacy office software to G Suite, giving users a suite of intelligent, cloud-based apps with a lightweight, modern interface. The team now collaborates across devices using Gmail, Google Docs, Drive, Sheets, and Slides, and uses Google Hangouts Meet for videoconferencing.

“All of our employees, including our CEO, were unhappy with our old collaboration tools,” says Michael. “When we moved to G Suite, the complaints stopped right away. Needless to say, the move was well received. I have never heard another complaint about our collaboration toolset. In IT, that’s the highest form of compliment.”

Modeling global finance

Since moving to Google Cloud Platform, Alpha Vertex’s compute requirements have grown fivefold. The company has scaled easily while keeping headcount flat, and its monthly hosting bill still has not exceeded the amount it was paying previously for dedicated bare-metal servers. With a high-performance, scalable infrastructure, Alpha Vertex can try new ideas using more nuanced and complex models, and eventually will be able to extend its service to all global markets.

“By choosing Google Cloud Platform, we can remain a small business and still be global and cutting edge,” says Mutisya. “As Google continues to provide us with early access to best-in-class technology such as Intel Xeon Skylake processors to run live alpha testing at scale, we can innovate even faster.”

About Alpha Vertex

Alpha Vertex builds cognitive technologies to change the way finance professionals discover and analyze investment opportunities, combining artificial intelligence with human insight to identify hidden relationships in financial information and drive better investments.

Industries: Financial Services
Location: United States
Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE
Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE