Today, Zapia readily supports complex requests and serves up insightful recommendations. For example, someone might ask ‘which smartphone is right for me?’ and Zapia can respond with various recommendations, where to find the best price, locations of both physical and online stores selling the recommended phone, and more.

Accomplishing this level of complexity and service for more than 2.5 million Latin Americans requires scalable, reliable infrastructure and large language models that can handle massive amounts of context in a multimodal, multilingual environment. After researching large language models, BrainLogic concluded that Anthropic’s Claude models on Google Cloud Vertex AI were the ideal solution.

Claude performed exceptionally well with Spanish and Portuguese natural language reasoning and was highly proficient in not just seeking answers from the right places, but learning from feedback to continuously improve how it handles requests. In addition, Claude on Vertex AI covered one of the most vital needs for BrainLogic: the ability to work in hyper-localized environments. Claude communicates with people in their unique local dialect and adeptly navigates cultural nuances and behavioral contexts, while Vertex AI maximizes reliability and reach across the region.

With Vertex AI, BrainLogic benefits from highly scalable, efficient infrastructure that keeps costs steady as the company grows rapidly. Vertex AI also provides advanced developer tooling to allow experimentation with models, robust security, data privacy, and compliance capabilities, and scalable, fully managed infrastructure to deploy Zapia to millions of users.