iArxiv

Redesigning a Machine-Learning-Driven scientific paper discovery tool

Role: UX/UI Designer ∙ Product Designer ∙ Illustrator
Client: arXiv - Machine-Learning-Driven Scientific Paper Discovery
Timeline: May 2022 - Feb 2023

Project Overview

iArxiv is a platform created by a group of engineering students collaborating with scientific leaders in Argentina. Using Machine Learning, the system learns each scientist’s preferred research topics and automatically sorts daily arXiv.org papers based on individual interests.

They brought us in to elevate usability, remove friction, and bring clarity to a tool used by researchers multiple times a week, without losing the simplicity and minimalism valued by the scientific community.

My role combined UX/UI Design, consulting, and visual direction, from research to UI and illustration.

Problem

Despite the value of the ML engine behind it, the platform’s experience fell short:

  • No onboarding or guidance

  • Overly complex filters

  • Confusing settings

  • Paper overviews with insufficient (or overwhelming) information

  • Visual design didn’t reflect the seriousness or credibility of the tool’s purpose


The result: users struggled to stay updated and to find relevant research efficiently, even though that was the core promise of the product.

Insights

To understand user patterns, we conducted a short survey (15 respondents) covering usage frequency, habits, and pain points. Despite the small sample, the findings were consistent across scientists.

Key insights were:

➤ Researchers use iArxiv 2 to 4 times per week to quickly scan new papers.

➤ Filters were the biggest friction, described as complex and difficult to use.

➤ Updates matter: users wante a mix of their main area + related domains.

➤ Settings were valuable but unclear, leading to trial-and-error.

➤ Paper overview is the determining factor for deciding whether to read more.

Solution

We redesigned the desktop and mobile experience to be cleaner, faster and easier to personalize, while keeping a minimalist, research-friendly aesthetic.

➤ Simplified filtering.
Filters are essential for daily use but were previously overwhelming. We reorganized them into logical, clear groups (date, score, category) and designed a smooth drawer interaction that reduces cognitive effort and decision time.

➤ Smarter, more discoverable settings.
Users can now choose preferred categories, adjust scoring, decide whether to show abstracts and configure email updates, and create instant personalization.

➤ More actionable paper overviews.
We designed multiple levels of overview clarity depending on what scientists valued in order to speed up scanning, the core use case.

Branding & UI

We refreshed iArxiv identity keeping a light, minimal and trustworthy tone.

The visual direction blends scientific clarity with digital simplicity, supported by custom illustrations and a calm, research-friendly color palette.

Wireframes

Outcome

The final experience delivers a faster, more comfortable reading flow tailored to scientists’ weekly routines.

By clarifying information hierarchy, simplifying filters, and refining settings, we reduced friction across the journey from scanning to deep reading. Cleaner metadata, intuitive categorization, and a minimal, friendly visual language work together to support focus, continuity, and ease of use, turning a dense research platform into a clear, readable, and dependable tool.


Live version: https://iarxiv.vercel.app/
Legacy: https://iarxiv.org/