Magellan AI Analyzes Podcasts and Scales to 600K Ads in 1 Year With Beam

Magellan AI is the all-in-one platform for podcast advertising intelligence, media planning, verification, and attribution. With the world's largest database of podcast advertising data, Magellan AI allows customers to understand advertising effectiveness, monitor competitors, and identify opportunities to scale advertising in the fast evolving audio space.

Challenge

Stability issues and scaling difficulties led to the search for a new provider

Previously, Magellan AI worked with two different cloud providers. Unfortunately, the team struggled with stability issues and difficulty scaling. In addition, simple infra tasks – like deploying new ML models – took disproportionately long. Dockerizing apps, setting up SSL certificates, and fine-tuning multiple services turned into a full-time maintenance job for the Data Science team.

Discovering Beam

Increasing throughput with Beam’s autoscaling

The tipping point came when Magellan AI’s CTO and team decided the operational headaches of their old setup was slowing down their product roadmap. After researching other solutions, Magellan AI turned to Beam for the ease-of-use and scalability they needed.

“What would have taken us literally days with any other provider we’ve tried took only hours [on Beam]” - Ryan Kerr, Senior Engineer at Magellan AI

Magellan AI was able to quickly expand from running only one ML model in production to running five. Beam’s one-click infrastructure enabled them to easily deploy new models to production. “Just put it in a wrapper and do beam deploy,” said Anastasia Simonoff, the Data Science Lead at Magellan. “It doesn’t get much simpler.”

Now, Magellan AI can easily scale up their deployment to handle large bursts in traffic, thanks to Beam’s autoscaling. Using Beam’s QueueDepthAutoscaler, Magellan AI has been able to scale to a 50% increase in throughput that otherwise would have required a full time DevOps engineer and a large GPU quota to achieve.

Results

Growing the business without worrying about infrastructure

Since migrating to Beam, Magellan AI has been able to grow the number of advertisements processed per month by over 50% – all while achieving strong reliability.

Beam’s infrastructure has been so effective at running Magellan AI’s tasks that the company has been able to dedicate less time managing their infrastructure.

The team no longer has to babysit containers or worry about daily restarts. “We’re getting nearly three nines of uptime,” said Ana. In addition, tasks like regenerating SSL certificates or manually managing Docker images were replaced by Beam’s automated processes. “We shouldn’t have to think about SSL certs,” Ryan emphasized. “With Beam, we don’t.”

For Magellan AI, adopting Beam enabled the team to focus on the core mission—analyzing podcast advertising data—rather than managing infrastructure. The ability to quickly deploy models, effortlessly scale to handle large bursts in traffic, and rely on 99.9% uptime has all contributed to faster innovation and a better experience for Magellan AI’s customers.