As AI adoption matured within the World Bank, the need emerged for a more practical, hands-on layer that goes beyond guidelines and learning. While the AI Portal centralized knowledge, the Model Garden & Playground was designed to operationalize it, enabling employees to directly interact with AI models and APIs.
The project focused on lowering the barrier to experimentation, helping both technical and non-technical users understand how AI can be applied in their workflows through real-time interaction.

Main objectives of this project were to:

• Provide a unified interface to discover and access curated AI models and APIs.
• Enable real-time experimentation with AI through an interactive playground environment.
• Simplify the process of testing use cases without requiring deep technical setup.
• Support responsible AI usage by embedding guidelines and guardrails into the experience.
• Bridge the gap between AI knowledge and practical implementation across teams.

As a core designer on the AI API Model Garden & Playground, I collaborated closely with Product, Engineering, and AI teams to translate complex technical capabilities into an intuitive and accessible user experience.
My role focused on designing a system that balances flexibility for advanced users with clarity for beginners. I developed user flows and interaction patterns that support model discovery, API exploration, and prompt-based experimentation.
I created wireframes and high-fidelity prototypes to define the playground experience — including input/output interactions, parameter controls, and response visualization. Special attention was given to reducing cognitive load while still exposing meaningful model capabilities.
Additionally, I worked on structuring the Model Garden, organizing models, metadata, and use cases in a way that supports exploration and comparison. I also contributed to aligning the design with existing platform standards to ensure consistency across the broader AI ecosystem.

The project followed an iterative, experiment-driven design approach, reflecting the exploratory nature of the product itself.
We began by identifying key user groups, from developers to policy specialists and mapping their needs when interacting with AI models. This informed the design of different entry points into the system, from guided exploration to more advanced API usage.
Through rapid prototyping and internal testing, we refined interaction patterns for the playground, ensuring that users could easily input prompts, adjust parameters, and interpret model outputs.
Close collaboration with engineers was essential to accurately represent model behavior and technical constraints in the UI. Feedback loops from demos and user sessions helped continuously improve usability and clarity.
Agile practices allowed us to iterate quickly, testing assumptions and refining features in short cycles while maintaining alignment with broader platform goals.

The AI API Model Garden & Playground significantly enhanced the World Bank’s AI ecosystem by introducing a practical layer for experimentation and implementation.
Employees were able to directly engage with AI models, leading to increased confidence in using AI tools and a faster transition from concept to application.
The playground experience reduced the friction typically associated with API usage, making AI more accessible to a wider audience beyond technical teams.
By combining model discovery with real-time interaction, the platform encouraged exploration and innovation, resulting in new AI-driven ideas and use cases across departments.

• The integration with the broader AI Portal ecosystem ensured a seamless journey from learning to doing — strengthening overall AI adoption within the organization.

Designing for Dual Audiences:
I learned to design for both technical and non-technical users simultaneously balancing depth with simplicity in a single interface.
Making AI Tangible:
Transforming abstract AI capabilities into interactive, understandable experiences became a core design challenge and growth area.
Interaction over Information:
Unlike content-heavy platforms, this project emphasized interaction design where clarity of inputs, outputs, and feedback loops is critical.
Reducing Complexity:
I strengthened my ability to simplify complex systems without oversimplifying them — preserving power while improving usability.
Cross-Functional Collaboration:
Working closely with AI engineers deepened my understanding of model behavior, constraints, and opportunities directly influencing better design decisions.
Designing for Experimentation:
I learned how to create environments that encourage safe exploration, iteration, and curiosity essential for AI-driven products.