Loka’s GenAI Work is Solving Celebrity Choice Conundrum
Industry
Media, Entertainment,
Machine Learning
Tech & TOOLS
AI/ML, AWS Bedrock,
AWS SageMaker
Teams & Services
Generative AI Engineers,
AI/ML Engineers
milestones
Narrowing 27 use cases to one POC Moving POC to Production
Industry
Media, Entertainment,
Machine Learning
Tech & TOOLS
AI/ML, AWS Bedrock,
AWS SageMaker
Teams & Services
Generative AI Engineers,
AI/ML Engineers
milestones
Narrowing 27 use cases to one POC
Moving POC to Production
Cameo pairs users with the perfect celebrity to deliver custom messages to friends and family. There’s only one sticking point: Choosing one option out of 50,000.
The situation
With 50,000 celebrities and creators available—from Ice-T to Brett Favre, the Naked Cowboy to Mr. Wonderful—Cameo’s visitors faced a daunting task in picking the right person to deliver their messages. Cameo wanted to streamline user search on their platform using generative AI on AWS, and they came to Loka for help.
The challenge
Loka’s aim was to help the average Cameo customer spend less time browsing and more time creating. A faster path to the perfect choice would improve Cameo’s UX, deliver greater customer satisfaction and increase repeat use. Cameo was already experimenting with AI—including Cameo kids, AI models trained by voice actors—but to arrive at a use case viable for search they needed Loka’s expertise.
The solution
Cameo partnered with Loka through our GenAI Accelerator program. Within the initial discovery phase with Cameo, Loka narrowed 27 possible GenAI use cases to one. The most viable was a use case that Cameo could afford to run in production and would directly improve customer experience. It also demanded GenAI expertise beyond their in-house team’s abilities. Loka’s engineers stepped in.
Over the course of the engagement, Loka identified the AI model and query prompts that most effectively refine user searches, narrowing choices based on query responses like city of residence, interests, budget and other variables. By conducting a GenAI-driven conversational experience with users, Cameo could match them to the most appropriate categories for choosing talent. The result was a new, viable user feature that Cameo could confidently bring to users, improving UX and shortening time to purchase.
“Our partnership with Loka allows us to move faster and punch above our weight class with more complex LLM implementations—the kind that remove friction and keep our customers coming back for more.”
What we delivered
Loka developed a POC that successfully integrated diverse data sources for a comprehensive and cohesive AI model training process and an AI solution capable of handling increased user activity and data volume. We also implemented stringent data privacy and security measures to ensure Cameo complies with regulations and safeguards user data.
The POC focused on model selection and prompt engineering, continuously aligning the AI model's recommendations and responses with user expectations to enhance the overall user satisfaction. We’re now deploying the model on Bedrock and supporting infrastructure (i.e., embeddings, vector databases) within Cameo’s AWS environment for testing.
These deliverables went beyond the theoretical or internal. With this viable, production-ready POC Loka applied GenAI to personalize and simplify the search process.
Project highlights
From 27 GenAI Use Cases to 1
Beyond cataloging use cases, within weeks Loka guided Cameo to a viable case study they could afford to run in production in the near and long term
From POC to Production
Loka is integrating their GenAI-powered solution to improve the search functionality
Streamlined AI Model Evaluation & Development
Evaluated suitable LLMs and determined effective query prompts, ultimately setting up Amazon Bedrock infrastructure in Cameo’s AWS environment, including embeddings and vector databases
Locktight Data Privacy and Security
Implemented stringent data privacy and security measures, ensuring compliance with regulations and safeguarding user data throughout the project and into production.
