Building trust in mobile communications

Struggling to achieve business value with legacy data tools

Today’s bad actors evolve quickly, finding new and sophisticated strategies for disrupting mobile device communications. First Orion protects consumers by accurately labeling unknown numbers as “Scam Likely” or with business details so people can engage safely and with confidence. Unfortunately, as their data volumes increased, First Orion’s legacy reliance on AWS EMR and other native services became overwhelming to manage, fundamentally slowing their data teams and preventing the organization from driving innovation based on insights and analytics.

“Our strategy was for the core data team to become the organization’s center of excellence, but we had so many data silos, duplicate data, new projects and requirements each week and complex infrastructure to manage that scaling was expensive and not agile,” says Zach King, Associate Director of Data Architecture at First Orion.

Building, managing, and maintaining clusters, notebooks, and production jobs was arduous and expensive. First Orion had to combine EMR and Jupyter notebooks and manually configure Apache Spark™ settings on a case-by-case basis. Clusters took one to six hours to provision, and standing up additional clusters was too labor-intensive to justify. Jobs were cued, productivity stalled, and data scientists had to read data in and write data out of notebooks constantly. To meet the peak resource requirements of workloads for data science and analysis, clusters ran 24/7 — leading to over-provisioned compute and operational costs as high as an average of $6,000 per user per month. This was untenable as data volumes continued to soar.

Furthermore, granting secure access to data under telecom compliance standards and ISO certification was a full-time job. The team resorted to anonymizing certain roles in data restores, which exploded the size of their data sets and ultimately the cost to manage them. To operate as a center of excellence — distributing valuable analytics and insights to business users for product development optimization — First Orion needed a modern architecture for data, analytics and AI.

Lakehouse architecture simplifies data use with operational efficiencies

Realizing that their current infrastructure needed an overhaul, they moved directly to Databricks Data Intelligence Platform for its unified approach to data, analytics and AI. With a managed platform on a simple infrastructure, First Orion has reduced operational costs through automated customer management, support for multiple programming languages, shared collaborative notebook environments, and easy access controls and governance.

Nysia George, Director of Data Science at First Orion, explains, “Databricks’ interactive clusters allow us to power machines more efficiently. We can share, interact and collaborate without any hindrance. Even as a director, I can navigate through team notebooks and easily get the insights I need.” Leveraging the user-friendly interface and Python, R, SQL and Scala integrations, data teams were effective immediately, producing advanced visualizations for business users and collaborating within consolidated notebooks.

As a result, over 200 First Orion employees across more than eight IT, marketing and sales teams are leveraging Databricks Data Intelligence Platform to run analytical reports, train ML models, automate billing, and build hundreds of ETL pipelines. Data teams have embraced 100% of the platform, including Delta Lake as the underlying storage layer powering reliable data pipelines, Databricks SQL for BI workloads at scale, MLflow for managing the ML lifecycle, and Unity Catalog for data governance, access control, auditing, lineage, and data discovery. Now that First Orion can dynamically manage data and AI, grant compliant access, and collaborate across teams, it’s jump-starting data-driven product development at an unprecedented speed.

Modernization delivers operational efficiencies that drive innovation

Since migrating to Databricks Data Intelligence Platform, First Orion has fulfilled a variety of use cases supporting an informed and secure mobile experience. First Orion’s INFORM® and ENGAGE® branded communication solutions identify legitimate businesses’ and brands’ outbound phone calls with the business’s name, reason for calling and logo. Built on lakehouse architecture, branded communication relies on anomaly detection, monitoring, fraud prevention and brand personalization to raise answer rates by 200%. Additionally, INFORM increases first-call conversion rates by 76%.

With SENTRY™, First Orion empowers businesses to spoof-proof their outbound numbers to prevent bad actors from scamming consumers. SENTRY uses a call protection API integrated with Databricks for call analytics and billing. It automatically authenticates carrier requests to ensure that calls from the business’s originating number can only be delivered to the destination during a finite time window. SENTRY helps brands avoid the costs of fraud resolution and brand damage while increasing agent satisfaction and productivity.

Internally, First Orion is maximizing the operational efficiencies of a unified platform and centralized data for impactful product developments. On Databricks Data Intelligence Platform, First Orion has boosted productivity by more than 35%. As a result, operational efficiencies and new product innovations have contributed more than $12.7 million in value over three years. King says, “From developers to architects to data scientists, we’re saving a massive amount of time now that they can focus on outcomes, collaborate on the same platform, easily access data, and use the tools to help other teams.”

Moving forward, First Orion will continue using Databricks Data Intelligence Platform to improve mobile communications with accessible data and AI. The team is using open models such as MosaicML MPT and Dolly on the lakehouse, creating large language models (LLMs) to replace manual document search in an effort to boost employee productivity, with preliminary results showing a 5x increase in search efficiency. Other LLM-based services target AI answering and scam interception. George says, “Team members can be independent without being siloed because Databricks has straightforward tools. Almost anyone can access and apply the insights and analytics we need to optimize communications.”

Utilizing the latest generations of LLMs such as MPT-7B, a 7 billion parameter model from MosaicML, recently acquired by Databricks, First Orion is looking to enrich SMS and phone calls with conversational AI. These LLMs would provide AI value that improves quality of fraud detection in honeypots, improve the speed and accuracy of entity vetting with human-in-the-loop workflows, and potentially add new branded communication offerings.

Unwavering in their commitment to provide secure and trustworthy mobile communications, First Orion’s commitment to lakehouse architecture positions them for continued success and transformative growth.