United StatesTechnology and telecom2025 – ongoingYext
Standardised AI model deployment on BentoML
Yext runs more than 150 models in production on BentoML and says development time fell 70%
Problem
Yext's data science team needed a common way to deploy many kinds of models for its products. The models ranged from generative and fine-tuned small language models to classical statistical ML. Yext also needed to serve them across several clouds and regions.
Approach
Yext standardised on BentoML's BentoCloud inference platform. The data science team deployed and scaled models without waiting on engineering. Multi-cloud deployment let Yext meet local rules and use GPUs with better rates and availability.
Results
Takeaway
A shared deployment platform let the data science team ship models on its own. All figures come from the platform vendor.
Categories
- Industry
- Technology and telecom
- Business function
- IT and R&D
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and MLGenerative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
BentoML's customer story says Yext serves generative, fine-tuned and classical models on BentoCloud across several clouds. It reports a 70% cut in development time, compute costs down by up to 80%, twice as many models shipped and over 150 models in production.
Full details are in the original sources.