AI & Systems

Context-Aware AI Walking Meditation (Smart Glasses)

Designed an AI-driven walking meditation experience that dynamically adapts guidance based on environmental context detected through smart glasses.

  • Contextual prompt architecture: structured prompt engineering translating environmental signals into psychologically coherent guidance
  • Latency-aware experience design: interaction pacing and response framing designed to accommodate real-world model delays
  • Hallucination mitigation through design: constrained prompting, environmental grounding and structured output formats
  • Evaluation programme: large-scale UX testing and comparative benchmarking of latency and contextual accuracy against baseline AI models

Conceptual diagram of context-aware AI meditation system

Context-aware AI pipeline translating environmental signals into real-time guided meditation.

AI-enabled Knowledge Base

Developed a design tool integrating academic literature, internal experimental findings and structured LLM reasoning into a unified knowledge layer for product teams.

  • Multi-source synthesis: research evidence, internal data and model reasoning combined into structured outputs
  • Accuracy controls: constrained prompting and evidence grounding to reduce hallucination risk
  • Design translation layer: converts psychological principles into actionable product recommendations
  • Team-aware guidance: suggests feasible designs aligned to available developer and artist skill sets

Conceptual diagram of AI knowledge synthesis and design recommendation system

Research evidence, internal data and LLM reasoning integrated into structured design intelligence.

Categorisation Engine

Built a system to automatically analyse and tag VR experiences at scale, turning design features into structured, analytical data.

  • Design dimensions tracked: colour, visual complexity, cognitive load, music, pacing, scene changes, interactivity and more
  • Automated analysis: extracts visual and audio features across large content libraries
  • Consistent framework: standardised categories allowing direct comparison between experiences
  • Scalable system: updates automatically as new experiences are added
  • Connected to user data: links design features to engagement and performance metrics

Conceptual diagram of automated VR experience categorisation system

Automated classification of immersive design features into structured data.

Real-Time Voice Analysis for Soft Skills Training

Contributed to the development of a real-time voice analysis system embedded within soft skills training applications, enabling users to receive live feedback on their communication performance.

  • Emotional voice analysis: detection of vocal cues linked to a person's emotional state
  • Sentiment analysis: speech-to-text conversion followed by natural language sentiment evaluation
  • Live performance feedback: real-time insights delivered during presentations and difficult conversation simulations
  • Behavioural integration: feedback designed to encourage immediate performance adjustment and skill refinement

Conceptual diagram of real-time voice analysis system within soft skills training

Real-time emotional and sentiment analysis integrated into behavioural training environments.