Gia – Product design, Figma, Design system

Evaluation of advanced language and multimodal AI systems to improve model reliability, safety and the quality of human-centred AI interactions.

• Evaluated advanced language and multimodal AI systems through structured testing, adversarial scenarios and rubric-based assessments;
• Identified reasoning failures, hallucinations and instruction-following issues;
• Turned findings into actionable insights for model reliability, safety and human-centred AI interactions.

An AI-native SaaS platform that turns customer conversations into searchable product intelligence. Product teams use it to find patterns across interviews and calls, pull out supporting evidence and make better-informed product decisions.

• Co-founded Chatty Fox and built the product end to end, from UX to LLM-powered features;
• Designed and built LLM-powered transcription analysis and structured information extraction from customer conversations;
• Designed AI-assisted workflows that help product teams spot patterns and surface evidence for decisions.

An independent product design and development practice serving smaller SaaS businesses and building its own AI products. The work covers AI-native SaaS, conversational tools, personalized retail experiences, iOS apps and production websites, taken from early concept to shipped product.

• Designed and built Style IQ, an AI fashion product: structured the catalogue, designed recommendation workflows and built the React frontend;
• Moved Style IQ from a rules-based approach to an agentic one with strict validation and feedback loops that keep AI outputs consistent with reliable product data;
• Led product strategy, UX, interface design and implementation for client products, from early concept to shipped product;
• Built internal design tooling with AI, including a plugin that exports design tokens to code and a Figma MCP + Codex workflow that turns designs into prototypes using design-system components;
• Worked across frontend, backend, APIs and deployment, designing for traceability, understandable AI outputs and private data handling.

A suite of AI-powered mineral exploration software built by a space-tech company and used by exploration teams worldwide. It turns dense geoscience data into decisions about where to explore. The products bring complex scientific analysis, 3D spatial data and AI-driven insights into one workflow, so geoscientists can review results and trust the evidence behind them. They are designed for specialists with deep domain knowledge but limited time for complex tools.

• Drove product design for Fleet’s AI-powered scientific software, reporting directly to the CTO and turning complex geoscience workflows into production software used by exploration teams worldwide;
• Initiated a shared design system across ExoSphere, ExoCore and Comet™, built on semantic tokens and reusable interaction patterns that translate directly into code;
• Partnered with geophysicists and engineers to revisit flows as scientific requirements evolved, keeping reusable patterns intact;
• Introduced progressive disclosure for geoscience visualisations, balancing stakeholder requests for simplicity with the evidence users need to trust the results;
• Prototyped AI-driven interactions in React with real data to surface issues static designs missed;
• Designed onboarding and workflow improvements that helped new users become productive faster with complex analytical tools.