CASE STUDY · AI RESEARCH WORKSPACE

Synthesis

An AI-powered research workspace that helps students and researchers upload, organize, and synthesize insights across multiple documents — replacing scattered reading with structured, connected understanding.

RoleUX/UI Designer

Duration3 weeks

TypeWeb App

ToolFigma

View Figma prototype ↗

Overview & problem

Students and researchers struggle to connect ideas across multiple papers, retain and organize insights from large volumes of content, and trust AI-generated summaries without knowing how reliable they are. Existing tools work one document at a time, summarize passively with no synthesis layer, and hide how the AI reaches its conclusions.

The goal was an AI-collaborative workspace that synthesizes research across documents while keeping the user in control of what they trust and act on.

Target users

Undergraduate & postgraduate students
Independent researchers with large reading lists
Knowledge workers synthesizing many sources

Design goal

Synthesis across documents, not single summaries
Transparent AI confidence on every insight
Keep the user in control of what to trust

User persona

Aditya Rao

24 · Postgraduate research student

Goals

Synthesize findings across 10–15 papers
Identify patterns and contradictions
Export structured notes for writing

Pain points

Reads papers one by one, no connecting layer
Highlights and notes scattered across tools
Can't tell reliable AI output from hallucination

User flow

Upload PDFs, paste URLs, or drop YouTube lecture links

AI processes and chunks the documents

View AI-generated clusters on the Synthesis Canvas

Ask follow-up questions in the Prompt Workspace

Review confidence indicators and citations on each insight

Export synthesized notes as a structured outline

Wireframes

Final UI

Usability testing

Tested with 3 users on an interactive Figma prototype
Added a legend and tooltip explaining the confidence scale
Introduced a default collapsed view for clusters
Added a document source tag on each cluster card
Made citation links visually distinct with an underline and icon

Key learnings

Confidence indicators are a design responsibility, not just engineering
Users need to feel in control of AI outputs, not just receive them
Transparent AI states (processing, chunking, ready) reduce anxiety
Citation linking is essential for high-stakes research tools
Early canvas wireframing prevented major structural rework

← Back to all work