AI recruiting platform lowers costs, improves performance via cloud repatriation

An AI-powered company repatriated from AWS to a private cloud to boost capacity and performance while cutting costs.

Challenge

A cloud-native AI recruiting platform reached maturity and found that the costs of its AWS infrastructure were becoming unsustainable. Initially attracted by the scalability of the public cloud, the company no longer needed many of the features it was paying for. What's more, scalability had become less of a concern: user demand and expectations were steady. Peak loads happened at predictable times. And because the company relied heavily on powerful graphics processing units (GPUs) to run its LLMs, it anticipated major cost increases as it continued to grow.

Solution

The company showed Summit its AWS bills and we knew right away that they could save money by repatriating to a private cloud. But before recommending the switch, we dove into its workloads, architectures, dependencies, performance patterns, and—most importantly—business context.

This analysis made it clear that the company could both save money and enjoy significant performance improvement by repatriating.

So we moved through the Summit Repatriation Framework (SRF) to ensure the process caused minimal business disruption, yielded fit-to-purpose solutions, and was de-risked at every stage.

Ultimately, the company repatriated to a private cloud operated by Summit for a monthly fee.

Results

• Cost savings: With Summit as its partner, the company spends 30% less on monthly cloud costs than it did with AWS.
• Increased capacity: When used to power a private cloud, the same budget paid to AWS delivered 50% more processing power than it did on the public cloud, meaning the company had plenty of room to grow without increasing spend.
• Human support: There’s no such thing as support from a hyperscaler like AWS. With Summit, the team can talk to a real engineer instead of an AI chatbot.
• Noticeable performance improvement: Swapping shared network storage in AWS for dedicated local storage from Summit led to a noticeable improvement on performance.

Conclusion

This company’s experience demonstrates how AI-powered companies that rely on GPUs to power LLMs can enjoy both cost savings and performance improvements by repatriating from the public cloud. This experience is particularly relevant to companies entering scale-up and those that have achieved a steady state. At that point, when workloads are stable and demand is predictable, the costs of the public cloud often start outweighing its benefits. Repatriating to a private cloud setup reduced the company’s operating costs and positioned it for growth.

However, it’s important to note that cloud repatriation should always be considered in the larger business context and, when it makes sense, executed with consideration for impact and future performance. The Summit Repatriation Framework takes those things into account to guide every business toward the hosting infrastructure that makes the most business sense.