Leaderboard charts for an AI evaluation company
An AI evaluation company had benchmark results in spreadsheets and wanted charts people would stop and share on LinkedIn and X. I turned the raw tables into a set of polished leaderboard visuals in their dark-and-gold style.
- Client
- AI evaluation company
- Made
- A pilot chart plus 4 visuals
- For
- LinkedIn and X
- When
- 2026
The problem
The company tests large language models and publishes how they compare. The results lived in raw data and Excel tables, which don't travel well on social media. They needed charts that make a complicated comparison readable in a few seconds and look good enough to share.
What I made
It started with a small pilot of one or two charts. Once that was approved, I made a set of four more visuals:
- Side-by-side rankings comparing models on two measures at once, such as safety and helpfulness.
- Scatter plots with a frontier line, showing which models give the best trade-off between two scores.
- Line charts of performance over time, showing where models start to slip.
- One consistent style: a dark slate background with gold picking out the top-ranked model, so every post is recognisable as theirs.
The result
- Raw benchmark tables turned into charts built for a social feed, readable at phone size.
- A visual style the team can keep using for future results.
Got great data that nobody reads? Book a free rant session.