Model Hub / Project "Next"
08/ A new, standalone application to simplify how users discover, learn, and manage complex AI models.
Company / Client
Distinct AI
Year
2025 - 2026
Type of Work
Vision
Product Roadmap
Product Strategy
User Research
Experience Design
Model Hub is designed and built from scratch to address four key technical and usability pain points:
1. Hardware Compatibility: To solve confusion regarding hardware specifications (GPU, CPU, VRAM), a custom-built 'Recommendation Engine' automatically detects the user's hardware and filters models to match their system's capabilities.
2. Visual Guidance for AI Model Suitability: Card-based presentation provides an intuitive UI that guides users to discover and learn about an AI model's capabilities without being overburdened by the technical complexity of variant quantities and hardware compatibility.
3. Download Management: Recognizing that model files can be massive (exceeding 30GB), the Hub handles downloads in the background. It provides real-time progress updates and toast notifications upon completion, allowing users to continue their work uninterrupted.
4. System Organization: The Hub acts as a blueprint for modular design, helping users manage an array of models across different applications and workflows without feeling overburdened by information
A prototype representation of Model Hub's Discover Tab during design and development. Design components are scaled from our newly built, in-progress design system shared amongst all "Next" era codebase to improve lead time.
Visually engaging "model cards" within the Discover Tab display key information at a glance, allowing users to quickly identify which models are most suitable for their specific hardware and output needs.
Key info includes:
-
Model variant name
-
Model capabilities, benefits
-
Simplified hardware details that match model to user's hardware
An in-progress schematic to illustrate the anatomy of the Model Hub's Discover Tab and how the "model cards" fit as a modular part of a system. Many iterations and variations were explored to solve both user and technical pain points.
A highly simplified sample workflow diagram to capture design and development intent. An important design philosophy of mine is to preserve optionality and to minimize design and technical debt.
Our teams have carefully considered tradeoffs between technology, platform, hardware, as well as end-user needs.
Download Management: Recognizing that model files can exceed 30GB, the Inventory Tab handles downloads in the background.
It provides real-time progress updates and toast notifications upon completion, allowing users to continue their work uninterrupted.
System Organization:
The Tab also helps users manage an array of models across different applications and workflows by
tracking versions, updates, and the status of downloads, installations, or uninstalls.
Team Effort
A Shout-Out To Those Who Knows
All of the behind-the-scenes heavy lifting was done by small teams in a pre-Claude Code era. Pixel by pixel and line by line of code, we pushed forward one checklist item at a time to build the Model Hub, along with every component and piece of IP across the Distinct AI ecosystem. IYKYK.
*Content in this page has been modified. All content shown is pre-production and is not intended to generate profit.