Product Strategist, UX/UI Designer
Latent AI
Catalyst is a web-based interface developed to enhance the Latent AI Efficiency Inference Platform (LEIP), an API-driven solution. This platform empowers users to assess machine learning model performance on specific hardware and facilitates the building and deployment of fine-tuned models to edge devices. Catalyst was designed to streamline the user experience, making the process of experimenting with and deploying machine learning models more efficient and intuitive.
Hardware manufacturers with technical expertise faced a significant challenge: the process of experimenting with, fine-tuning, building, and deploying machine learning models was complex and often required extensive coding. consuming, and presented a barrier to efficient AI integration. The need was clear: a user-friendly interface was required to simplify and streamline the workflow, enabling these manufacturers to seamlessly incorporate AI into their hardware. This process was cumbersome, time-consuming, and presented a barrier to efficient AI integration. The need was clear: a user-friendly interface was required to simplify and streamline the workflow, enabling these manufacturers to seamlessly incorporate AI into their hardware.
The primary users of Catalyst are technically proficient hardware manufacturers seeking to integrate AI into their product offerings. These manufacturers often specialize in camera-equipped hardware, including drones, security systems, and specialized industrial applications like bowling alley score detection. These users may or may not possess in-house software engineering capabilities.
As Product Strategist and UX/UI Designer, I led the UI design for Catalyst and collaborated on product strategy with one other designer.
Collaborating on product strategy and roadmap
Collaborating with engineers to ensure technical feasibility and successful implementation.
Collaborating with engineers to ensure technical feasibility and successful implementation.
Iterating on the design based on user feedback and testing
Users begin by creating a new project.
Users select their target hardware.
Users experiment with pre-tested hardware/model combinations ("recipes") to evaluate performance. "Recipes" provide a starting point, streamlining the optimization process.
Users build the model.
Users deploy the model to their hardware.
Upon logging in, users are presented with a clear overview of their previous work and the current status of ongoing projects (compilation or training). This provides immediate context and allows users to quickly resume their work.
The "Recipes and Hardware" screen provides a visual comparison of different hardware and recipe combinations. Key performance metrics, such as inference time, are plotted, enabling users to quickly assess trade-offs and identify optimal pairings for their specific use case.
The "Datasets" screen provides a clear overview of uploaded and Latent AI datasets, allowing users to easily manage and select data for their projects. Key information such as dataset names and validation status is displayed, and users can preview datasets and create new ones.
Catalyst empowers users to rapidly deploy machine learning models to edge devices with minimal coding. Its clear navigation, contextual information, and visual comparison tools enable faster decision-making, improved model performance, and significant time and cost savings. Simplified dataset management and transparent job feedback further enhance user productivity and satisfaction.
By streamlining the deployment of machine learning models to edge devices, Catalyst offers Latent AI several key business advantages. It expands the company's market reach by attracting hardware manufacturers with varying levels of software expertise, and reduces support costs by simplifying the integration process. The platform's intuitive design enhances user satisfaction, leading to increased adoption and positive word-of-mouth. Ultimately, Catalyst positions Latent AI as a leader in edge AI solutions, driving revenue growth and market share.