Crypto.com provides precise mood analyzes in 1 second with the help of generative AWS in AWS
Crypto.com will be carried out on Amazon Web Services (AWS) right from the start. When the company was looking for new LLMS for mood analysis, it was Anthropic Claude 3 in Amazon Bedrock A logical choice. These LLMs are highly scalable and process large amounts of data in real time, which enables comprehensive market research. In the first tests, it was shown that LLMS in Amazon Dark will deliver results very quickly, usually within a second.
By integrating Amazon Dampf, the manual effort, the additional costs and the calculation restrictions, which are associated with self -hosted LLMs. Within a month, Crypto.com implemented the Claude 3 Haiku models from Anthropic for mood analysis in Amazon Bedrock. This collects and analyzed crypton messages in more than 25 languages. The company also uses Amazon Redock for ongoing Proofs of Concepts (POCs) and for development. Sunny Fok, Head of Ai and Innovation Technology at Crypto.com, says: “The generative KI in AWS services such as Amazon Sagemaker and Amazon Rolle has optimized our use of the latest LLMs and AI technologies. Now we can further develop innovative ideas from the Proof of Concept within a few weeks.”
In order for specialist knowledge to flow into the output of the models, Crypto.com used its own data to optimize open source models such as Mistral you have and Goal call in Amazon Elastic Compute Cloud (Amazon EC2). This procedure is crucial when new coins come onto the market, since finished models often only provide insufficient results. Then crypto.com started, Amazon SageMaker to use as a need-controlled platform for complete ML development to optimize its tailor-made models.
Raymond Lam, Senior Engineer at Crypto.com, says: “Amazon Sagemaker offers the tools and APIs to effortlessly adapt our models with a user-friendly surface. Similar to Amazon, we only have to carry out machine learning orders as required. And it is much easier to manage tailor-made models than to do everything yourself. “
Crypto.com received comprehensive support from AWS when it comes to multi-agent provision in the phases. “We had several conversations about which tools or frameworks should be used for the various applications,” said Lam. “The AWS team informed us about existing applications, provided example code and demonstrated the individual steps. The AWS solution architects appeared as technical problems in the troubleshooting and submitted corresponding suggestions, which accelerated our incorporation into the generative AI in AWS services.”
