AIGF
Role
Senior Product Designer
Platform
Web — Start-up Environment
Startup MVP - Metrics & Impact: Delivered the MVP on time, helping build investor confidence and contributing to a multi-million-euro Spanish Government investment to further develop the platform and advance AI research. The Design System reduced design and development effort by 70%, accelerating delivery and enabling more consistent, scalable product experiences.
CONTEXT & ROLE
Led the Product Design team
Co-created the product in collaboration with Telefónica
Transformed complex technical requirements into intuitive workflows
Delivered the MVP within five months to support the Spanish Government investment process
AIGF (AI Gateway Framework) is an enterprise AI platform co-created by MCV, Multiverse Computing, and Telefónica, bringing AI resource management, deployment, monitoring, and experimentation into one unified experience. Combining Telefónica’s cloud and enterprise capabilities, MCV’s product and UX expertise, and Multiverse Computing’s AI technology, the platform provides organizations with a single access point to manage and scale AI services.
CHALLENGE
The challenge was to transform a powerful technical platform, originally built “by engineers, for engineers,” into a clear and intuitive experience. As the product evolved, new features and patterns had created inconsistent interfaces and fragmented workflows, making the platform increasingly complex to navigate.
We started by identifying our key user groups, understanding their workflows, and defining what each needed from the platform.
Infrastructure Operators manage GPU clusters and Kubernetes jobs. They need high information density, fast access to technical data, and a clear operational overview. While the platform contained valuable information, it was difficult to navigate and understand efficiently.
AI Workspace Users build AI solutions, explore models, create agents, and configure RAG workflows. Unlike Infrastructure Operators, they need to accomplish these tasks without deep infrastructure expertise, requiring a simpler and more guided experience.
The goal was to design a platform that could support both complex infrastructure management and AI creation, while making each workflow clear, efficient, and accessible to its intended users.
Approach
1. Structuring AIGF as one platform, not a collection of features.
Working with Product Managers and Engineers, I helped define the information architecture, user journeys, and core platform structure, mapping key workflows into a navigation system where each module represented a clear area of the platform, rather than a loose set of independently shipped tools.
2. Building a scalable, multi-brand Design System.
Rather than separate systems for Telefónica and Multiverse Computing, I led a shared foundation of components, patterns, and design tokens, letting each brand keep its own identity while ensuring the platform felt like one consistent product. This directly targeted the interface fragmentation that had accumulated as features shipped independently, and became the reusable foundation for every subsequent workflow.
Storybook
An open-source frontend tool for building, testing, and documenting UI components in isolation, outside of the actual application they belong to.
APPROACH
Shared Design System foundation
Flexible multi-brand token system
Separated styles from components
Switched themes without rebuilding
Standardized design foundations
Ensured consistency, scalability, and WCAG compliance
Reduced duplication and improved scalability
DEV COLLABORATION
Shared foundations for spacing, typography, and sizing
Same nomenclature conventions across teams
Reduced design and development process, in terms of time
Improved efficiency, and component reuse
Wireframes
After defining the information architecture and user flows, we moved into wireframing to validate the core concepts and workflows before investing in final visual design. We developed high-fidelity wireframes using the Design System library, applying a consistent set of components and patterns across the platform. The designs addressed the needs of both technical and non-technical users, creating clear workflows for AI resource management, deployment, monitoring, and experimentation within a unified experience. This stage helped us validate the product structure, identify gaps, and align the design with technical and business requirements. The goal was to bring these validated concepts together into the MVP, which was presented to stakeholders and investors to secure support and funding for the next stage of development.
Specific problem: Redesigning the Agent Builder to reduce configuration friction
The Agent Builder configuration form was one of the clearest friction points in the platform. It presented LLM defaults, backend setup, runtimes, and agent settings all at once, without clearly distinguishing required from optional fields or showing users where they were in the process. In usability sessions, users spent roughly 60% of their time interpreting unclear fields rather than configuring their agent. This highlighted a broader UX issue: a workflow originally designed for Infrastructure Operators was also being used by AI Workspace Users with different levels of technical knowledge and different needs. The result was a long, cognitively demanding form that created frustration and slowed task completion.
The solution
I redesigned the configuration experience around progressive disclosure, turning one long form into a guided, sequential workflow:
Progressive disclosure: Sections start collapsed and reveal information only when users need it.
Essential information first: The most important configuration options are surfaced first, reducing the initial cognitive load.
Smart defaults: Fields are pre-filled wherever a sensible default exists, reducing unnecessary decisions.
Persistent Summary: A summary panel shows completed and remaining steps, giving users a clear sense of progress.
Review & confirmation: Before submitting, users can review the entire configuration in one place, verify their choices, identify missing or incorrect information, and make changes before creating the agent.
Auto-save: Configuration is automatically saved, removing the need to copy information into external notes to avoid losing progress.
Impact
The redesign transformed the Agent Builder from a long, intimidating configuration form into a guided workflow that better supports users with different levels of technical expertise. The Review step also added a final validation point, giving users confidence that their configuration was complete and correct before submission. More importantly, the solution established a reusable interaction pattern for AI-creation workflows across the platform, providing a foundation for designing more guided and accessible experiences beyond the Agent Builder itself.
Outcome
The project transformed AIGF from a technically driven MVP into a scalable, user-centered platform supporting both Infrastructure Operators and AI Workspace Users.
MVP delivered in 5 months, supporting product validation and investment.
70% reduction in design and development effort through a reusable Design System.
Clearer, more guided workflows for complex AI and infrastructure tasks.
Reusable interaction patterns that created a foundation for future platform growth.