Automated Comet Weekly Experiment Scorecard
Every Monday, WebRun opens Comet, reviews experiments logged in the past week per project, exports the best run and its metric to a Google Sheet, and posts a scorecard page to Notion so the team can see progress without digging through the experiment list.
How do I get a weekly scorecard of my Comet experiments?
WebRun reviews Comet every Monday, finds the best-performing experiment run per project from the past week based on its tracked metric, and logs it to a Google Sheet alongside its parameters. It then posts a scorecard in Notion comparing this week's best result to last week's, so ML progress is visible without digging through raw experiment logs.
- The best result each week is visible without opening the experiment list
- Week-over-week progress is tracked automatically in one sheet
- The team gets a scorecard in Notion instead of hunting through Comet
Built for ML engineering teams · data science teams · research leads · MLOps teams
What does WebRun do on every run?
The exact actions WebRun takes, in order - in plain language, so you can adjust anything.
-
WebRun signs in and gets to work
Opens
www.comet.com/signinin a real browser with your saved login - no setup, no API keys. -
1
Comet - review the week's experiments
- Open Comet and review experiments logged in the past week per project
- Sort each project's experiments by its primary tracked metric
- Note the best run, its metric value, and key parameters
Done when The best experiment run of the week is identified per project.
-
2
Notion - export the best run per project
notion.so
WebRun opens Notion to post the weekly scorecard. - Open the ML team's Notion page
- Post a scorecard naming the best run per project and how it compares to last week
- Link to the run in Comet
Done when The team has this week's experiment scorecard in Notion.
-
3
Google Sheets - post the weekly scorecard
google.com
WebRun opens Google Sheets to export the best run per project. - Open the Comet Weekly Runs sheet
- Add a row per project with this week's best run, its metric, and parameters
- Keep prior weeks' rows for trend comparison
Done when The sheet has this week's best run logged per project.
How is each run configured?
Secure by default
Connect once, stays signed in
WebRun signs in once and keeps each session in a persistent environment, so every run picks up right where it left off.
Every action is checked against this policy before it runs.
Questions, answered
Does WebRun register or promote a model itself?
No. It only reports on the best run per project. Registering or promoting a model to production stays a decision for the ML team.
How does it pick the best run?
It uses the primary metric already configured for each project in Comet, so the ranking matches what your team already tracks.
What if a project had no new experiments that week?
That project is simply skipped in that week's scorecard, so the summary only covers projects with fresh results.
Related templates
Put this on autopilot.
Turn it on in minutes - or have our team set it up for you.