Get in touch

Harnessing the Power of Machine Learning & AI for User-Centric Solutions

Welcome to an exploration of how cutting-edge Machine Learning and AI technologies can be seamlessly integrated into user-centric designs. This case study demonstrates my approach to solving complex problems with innovative AI-driven solutions, ensuring a balance between technological advancement and exceptional user experience.

My role: Project Manager/Dialogue management/ UX testing and evaluation 

Project Overview

Transforming Data into Actionable Insights

In this project, I used Machine Learning and AI to create intelligent solutions that cater to user needs. The focus was on transforming raw data into actionable insights, enabling more informed and effective decision-making.

  • Objective: Develop AI-driven features that enhance user engagement and satisfaction by providing informative and accurate advice.
  • Scope: Integration of predictive analytics, recommendation systems, and natural language processing.

AI Model Training

Enhancing Emotional Intelligence with Rasa and Python

By training existing AI models with emotional intelligence, we aimed to create a more responsive and empathetic user experience. Utilising Rasa and Python, we meticulously fed datasets into these models to refine their performance.

  • Model Training with Emotional Intelligence: Training existing models, focusing on emotional intelligence to better understand and respond to user inputs.
  • Manual Data Feeding: Carefully curated and manually fed datasets into the models, ensuring they were trained with high-quality, relevant data
  • Validation: Employed rigorous validation techniques to ensure the models delivered reliable and valuable insights, maintaining high performance standards.

This approach ensured the AI models were not only technically robust but also capable of providing a nuanced and empathetic user experience.

This video demonstrates the emotional intelligence of our AI chatbot, built using RASA and PyCharm. These platforms enable us to train AI models and test applications using Python. As shown, the chatbot can decipher user inputs and respond with emotionally appropriate emojis or GIFs, demonstrating empathy.

User Research and Data Collection

Understanding the data

A thorough understanding of both the data and the users was crucial. I conducted extensive research to ensure the AI models were built on a foundation of accurate, relevant data.

  • User Interviews: Conducted detailed interviews with target audience to understand user pain points and accurate expectations of the system.
  • Surveys: Deployed comprehensive surveys to gather a broader spectrum of user feedback, ensuring diverse perspectives and needs were considered.

Dialogue management and response handling

The process begins with dialogue management, where the chatbot refers to its intents—tokenized chunks of text manually labelled from user interview scripts. If it cannot find a suitable response within the intents, it then scrapes information from a designated webpage. In this case, the chatbot accesses a private page containing information for program development tutors to formulate an appropriate response and display emotional intelligence.

Architecture Design

Response flow and back-up plan

If the chatbot is unable to return an appropriate response from the Rasa server, it defaults to the next option, which is to provide a generic response from Chat GPT-3.5 Turbo.

User Testing and Evaluation

Ensuring an Exceptional User Experience

I led the user testing and evaluation process, utilising proven methodologies to ensure our system met high usability standards.

  • System Usability Scale (SUS): To assess usability, I employed the System Usability Scale (SUS), a reliable tool for measuring user perception of the system. The SUS scores provided valuable insights into user satisfaction and areas needing improvement.
  • Identifying Improvement Areas: Based on the SUS scores, I pinpointed specific aspects of the system that required enhancement, guiding the refinement process.
  • Iterative Testing and Refinement: This iterative cycle of testing, refining, and retesting allowed us to progressively enhance the user experience. By continuously incorporating user feedback, we added more intents and useful information, improving the system’s responsiveness and accuracy.

This meticulous approach ensured the system evolved to meet user needs effectively, delivering a seamless and satisfying experience.

Results and Impact

Delivering Value

The AI-driven features significantly enhanced the user experience, providing personalised and predictive insights that empowered users.

  • Enhanced Engagement: Users reported increased satisfaction and engagement due to the personalised recommendations and insights.
  • Improved Decision-Making: The predictive analytics enabled users to make more informed decisions, adding tangible value to their experience