Machine Learning Collective

How might we redesign the site to grow funding, new publications, and the community at scale?

Machine Learning Collective

How might we redesign the site to grow funding, new publications, and the community at scale?

Machine Learning Collective

How might we redesign the site to grow funding, new publications, and the community at scale?

TEAM

Time

6 months

My role

Product Designer

KEY SKILLS

Growth Design

UX Research

AI Prototyping

Overview

I led the 0-1 redesign for MLC's responsive web experience to drive platform growth, using AI for rapid prototyping and build

MLC is a nonprofit aimed to democratize AI Research in providing a turn-key research platform for anyone, anwhere. The exisitng site made it hard for people to understand the value, discover opporutnities, or take action. The founders were looking to grow the platform, at scale.

SOLUTION

Research revealed three key value paths to growth - "join a project," "book researcher," and "attend an event"

I chose projects as the North Star Metric because it was the leading indicator for the ultimate goal, which is publications. Every action taken drives user either directly or indirectly to "Join or Submit a Project/Paper idea."

Tradeoffs: I deprioritized mentorship and events to focus the experience on projects - identified as the North Star Metric from research— chose clarity over comprehensiveness since users already felt overwhelmed by the existing experience.

product thinking

I redefined success around a North Star

User research surfaced three primary goals: attend events, talk to a researcher, and join a project. I prioritized "Join a Project" as the North Star Metric because it feeds directly into the publication pipeline. Publication being the end goal.

I chose Join a Project as the primary CTA because it directly feeds the publication pipeline.

who i designed for

Three personas, one North Star

Through user interviews and stakeholder interviews these personas surfaced. I mapped the most likely engagement loops for each persona.

research

I led with usability testing, not visuals, for the 0-1 redesign

Before designing anything, I ran moderated usability tests on the existing site with six independent researchers across my three personas — watching where they stalled, what they ignored, and what they'd actually come to do.

01 — WHAT USERS CAME TO DO
02 — WHAT I HEARD
83% of users dropped off in the first 3 seconds
83% of users dropped off in the first 3 seconds

“ The text on this site, it's too much and I never find the info I need. There's no clear way to find a collaborator. ” - Sam L.

100% of users wanted casual ways to start a research paper or join in on a research idea

“ I just joined Discord because it's not clear where I can find these researchers to talk to, or if there's events I can go. ” - Nick B.

100% of of users wanted to see the real researchers & papers first

“Honestly, this site is confusing. I mean, where are the researchers? Also the wording seems intimidating for submitting an idea.” - Joanna L.

100% of users wanted casual ways to start a research paper or join in on a research idea

“ I just joined Discord because it's not clear where I can find these researchers to talk to, or if there's events I can go. ” - Nick B.

100% of of users wanted to see the real researchers & papers first

“Honestly, this site is confusing. I mean, where are the researchers? Also the wording seems intimidating for submitting an idea.” - Joanna L.

key product decisions

A few examples of how research changed the proudct

CLEAR HERO AND VALUE

Proactively communicate MLC mission and value, clear NSM as primary action.

focused discovery

Discovery mapped to user goals making projects, events, and researchers easy to find.

Streamlined actions

Simplifies how users take action and get involved. 1 click actions from the homepage.

validation

Outcomes

Validation tested on the same six users, I measured the before and after for — time spent on page, how confident they felt in achieving their goal, and seven different tasks. Results validated behavioral change and optimization for growth.

Increased Time on Page 

+92%

User Confidence Score

+99%

Task completion rate

+6x