Inside this project
Making AI Development Intuitive & Action-Oriented
Microsoft AI Foundry is a platform for exploring, building, and deploying AI models and agents. When I joined, the platform had powerful capabilities—but the experience was fragmented and difficult to navigate.
My work focused on making AI development more intuitive, guided, and action-oriented—shifting the platform toward an AI-first, workflow-driven experience across multiple core surfaces.
Developer Challenges
Model Overload
Choosing between thousands of models without clear guidance
Configuration Complexity
Understanding requirements and translating ideas into implementations
Fragmented Tools
Navigating scattered tools and documentation across the platform
An introduction to the AI Foundry platform and its developer workflows.
Leveraging AI Design Tools
In addition to traditional design tools, I used AI-powered tools to accelerate ideation, prototyping, and implementation across these projects.
Figma
Primary design and prototyping tool
Figma Make
AI-powered rapid prototyping and iteration
GitHub Copilot
AI-assisted code generation and exploration
Microsoft AI Tools
Proprietary AI tooling in Visual Studio
Foundry Agent (Ask AI)
Problem
Developers were constantly leaving the product to search documentation, compare models, or troubleshoot issues.
Approach
Originated and drove an AI assistant from hackathon prototype to production-shipped feature—designed to answer platform questions, recommend models, and enable direct actions within the product.
The assistant becomes an entry point for platform questions and direct actions.
Exploring contextual assistance within a developer workflow.
Evolution
- Hackathon → Prototype: Built grounded Q&A assistant using internal docs
- Side Panel Assistant: Introduced conversational UI within the product
- Platform Integration: Became primary entry point via global search
- Agent Capabilities: Enabled actions like deploying models and managing quotas
- Production Shipped: Now a core part of the Azure AI Foundry platform experience
“AI shouldn't just live in a sidebar. It should be woven directly into the product experience.”Key insight
Responsible AI
Problem
Organizations needed to manage Responsible AI policies, content filters, and safety constraints across entire fleets of models and agents—but existing tools made this complex, fragmented, and difficult to scale across teams.
Approach
Designed comprehensive systems and UX patterns that enable developers, IT administrators, and program managers to configure, monitor, and govern AI safety at scale—from individual model guardrails to fleet-wide policy management.
Solution Highlights
- Template-based guardrail configurations for rapid, consistent policy deployment
- Fleet management interfaces for governing multiple models and agents simultaneously
- Clear organizational hierarchies that map safety controls to business units and use cases
- Role-based access patterns for developers, compliance teams, and administrators
- Unified visibility into safety postures across the entire AI deployment landscape
Impact
Transformed Responsible AI from a per-model configuration burden into a scalable, organization-wide governance system—reducing risk, improving compliance, and enabling safer AI deployments across enterprise environments.
Playgrounds
Problem
Developers needed to experiment across modalities (chat, image, video, voice, agents) with thousands of models from external providers like OpenAI, Meta, and DeepSeek. The challenge was compounded by tight deadlines to support their latest model APIs and features, driven by contractual obligations and competitive pressure.
Approach
Led design efforts to unify Playgrounds by creating consistent, scalable interaction patterns that work across all modalities and providers, supporting both rapid prototyping and deeper configuration while managing constant API changes.
A product walkthrough of the Foundry playground interface.
A product walkthrough of the Foundry playground interface.
A product walkthrough of the Foundry playground interface.
Key Contributions
- Designed end-to-end flows across multiple playground types (chat, image, video, audio, agents)
- Created scalable patterns to accommodate thousands of models from diverse providers
- Built flexible configuration systems to adapt to rapid API changes and new features
- Simplified complex parameter configuration while maintaining power-user capabilities
- Defined how developers move from experimentation → production deployment
Impact
Playgrounds became one of the most heavily used surfaces in Foundry, serving as the entry point for many developers exploring AI capabilities. Successfully scaled to support thousands of models while maintaining usability.
Project Templates
Problem
Even with powerful tools, many developers didn't know where to start or how to structure a real AI project.
Approach
Led design for Project Templates, providing pre-configured starting points that reduce setup complexity and bridge experimentation with real-world applications.
Design Principles
- Start from intent, not tools
- Reduce blank-page anxiety
- Guide without over-constraining
Impact
Templates became a critical onboarding surface, helping new users get started faster and experienced users accelerate common workflows.
Design System
Problem
As Foundry rapidly evolved with new features and surfaces, the lack of consistent, reusable components led to fragmentation and inefficiency.
Approach
Contributed to building and extending the design system with flexible, production-ready components that work across multiple surfaces and use cases.
Key Contributions
- Extended existing Microsoft design systems (including Fluent and Copilot patterns) into Foundry
- Created reusable components including Cards, Wizards, Draggable lists, and configuration panels
- Designed flexible component architectures that adapt to diverse product needs
- Improved consistency and efficiency across Playgrounds, Templates, Responsible AI, and Agent surfaces
Outcomes & Impact
These approximate figures describe platform-wide scale and shared team outcomes, rather than the isolated impact of an individual design change.
Monthly Active Users
Contributed to a platform used by hundreds of thousands of developers
Deployment Increase
Helped increase model deployments significantly
Core Surfaces
Designed foundational experiences across the platform
Scale & Governance
Enabled safe, compliant AI deployments for large organizations