AI & Analytics Engine

The AI & Analytics Engine (AI&AE) is a web-based machine learning platform built by PI.EXCHANGE that helps organisations quickly, easily, and affordably build predictive models and turn data into actionable decisions — without requiring deep coding expertise.

Purpose-built for anyone who works with data, the Engine guides users from raw dataset to deployed prediction model through a structured, simplified workflow. My responsibility spanned design sprints, usability testing, UI/UX design, and building the design system that unified the entire product experience.

My role involved conducting design sprints across various stages of the user journey — fostering collaboration between data scientists, engineers, product owners, sales, and marketing teams. To drive feature innovation, I facilitated workshops using affinity mapping and card sorting techniques.

I was also responsible for establishing and maintaining the AI&AE design system from scratch, ensuring cohesive visual and interaction standards across the entire product — and overseeing accessibility testing to make sure the experience worked for all users.

The design process followed a five-stage framework — grounding every decision in user understanding before moving to execution. Each stage fed into the next, with feedback loops built in to prevent design decisions from outpacing validated knowledge.

The research phase combined qualitative and quantitative techniques to build a grounded understanding of user needs, behaviors, and pain points. User research and brainstorming sessions used affinity mapping and card sorting — conducted in Miro — to organise findings into patterns that directly informed the product direction.

This included: conducting user interviews to surface goals and pain points; observing users in natural workflows; running surveys for quantitative signal; analysing existing analytics and customer service records; and creating user personas to represent the distinct types of users the platform served.

Research surfaced two primary personas with meaningfully different technical backgrounds, goals, and challenges — both of whom needed to find value in the same product. Designing for both without privileging one over the other was a core constraint of the entire project.

Before any screen design began, I mapped the complete user flow — from onboarding entry through to model deployment and export. This gave the team a shared understanding of every decision point, branching path, and moment where users could get lost or drop off. The flow directly shaped the simplified UX architecture that followed.

To create high-fidelity UI screens from wireframes, I followed a structured progression: adding colour, typography, and branding to give screens a professional finish; incorporating real content — text, images, and data — to make mockups accurate and stakeholder-ready; gathering user feedback and iterating; and adding animations, hover states, and micro-interactions to add realism and operational clarity.

Real-time collaboration tools kept the team and stakeholders aligned throughout — with every iteration reviewed and refined before moving to the next fidelity level.

The redesign was grounded in simplifying a technically complex product without removing the capability that expert users needed. Every major decision had to serve both personas simultaneously — the experienced data scientist and the aspiring analyst — without designing separate products for each.

The simplified UX divided the entire user journey into four major sections — each designed to be approachable on entry but capable at depth. Every stage surfaced the right controls for the right moment, without overwhelming users with complexity they hadn't yet needed.

The AI&AE visual design applied principles of contrast, hierarchy, and balance to create an aesthetically clear and easy-to-understand interface. Colour, typography, and imagery choices were grounded in understanding the audience — enterprise data teams and aspiring analysts — and the high-stakes nature of the decisions the product supported.