Berkeleytime · ML & Explore
OCTO Showcase · Spring 2026
Pod Showcase
Find the course you didn't
know to look for.
know to look for.
Berkeleytime · ML & Explore
The Solution
Our Approach
Don't rebuild the catalog.
Build something new beside it.
— Catalog · keep as-is
Look up what you know.
Fast
Simple
Informative
— Explore · brand new
Discover what you don't.
Personalized
Community
Discovery
Berkeleytime · ML & Explore
Phase 01
Phase 01
Pre-Midterm.
Two teams, moving in parallel.
— Design Track
Started with students.
→
— Engineering Track
Three problems. One pod.
01
Semantic Search
Search by meaning, not keywords.
02
Recommendation
Trace user behavior, surface adjacent courses.
03
Enrollment Probability
Predict P(get in) from waitlist position.
Berkeleytime · ML & Explore
Sub-team 01 · Semantic Search
Sub-team 01 · Semantic Search
From keywords to meaning.
She typed "how to take a vlog" — and BT had no idea.
— Before
→
— After
Berkeleytime · ML & Explore
Sub-team 02 · Recommendation
Sub-team 02 · Recommendation
"Because you viewed…"
Berkeleytime · ML & Explore
Sub-team 03 · Enrollment Probability
Sub-team 03 · Enrollment Probability · 3 Years of Data
What's my real chance of getting in?
— 3 years of enrollment data
→
model
— Your real chance, today
Berkeleytime · ML & Explore
Phase 02 · Integration
Phase 02
Post-Midterm.
From concept to shipped.
Figma
→
Live