Teaching and Running 6.390
AI as TA
Shen Shen
EECS
Engineering Council | March 30, 2026
What makes 6.390 Intro to Machine Learning an interesting case study
1. The content is fundamental, topical, and abundant online
2. The operation is complex: 350-450 students, 50-70 staff, many moving pieces
How to use AI to add value? → creative collaborator
Lots of routine work → delegate to scale
These are the clearest wins to me.
I:
Creative collaborator
Brainstorm for analogies
Strategize for design
critique and improve (iteratively)
Generate script
to smooth out segues
Vibe-code animations
Same concept, but I could afford to explore more because iteration was cheap.
2024, manually coded in 4 nights
2026, AI vibe coded in 30min over 10 designs
Inspiration from podcast
- Generate a podcast from my notes and slides
- The back-and-forth surfaced hand-off moments I wouldn't have thought of alone
- to change tones
- to invite questions or reflection, or
- to go on quick tangents
\(\theta^*=\left({X}^{\top} {X}\right)^{-1} {X}^{\top} {Y}\)
"Is Jane Street wrong about OLS formula?"
6.390 teaches:
AI allowed me to rewrite in our notation, land the provocative joke cleanly
Shirt uses a different notation. In old days, I'd abandon the idea.
Delivery coach: automatically critiques how I delivered, not just what I said.
A feedback loop no colleague or TA could practically provide.
| Lec05 | Lec06 | Change | |
|---|---|---|---|
| uh + um | 312 | 113 | -64% |
| "um" alone | 182 | 54 | -70% |
e.g. diagnosed when and why I use filler words, gave personalized tips, and continues to auto monitor Panopto recordings on schedule:
similarly for accent and grammar fixes (e.g. I never knew I used to pronounce `error` wrong...)
Lowered creative contribution barrier for everyone
Now:
- All TAs are contributing
- Lots of LAs (sophomores) are contributing
| Branches | Contributors | |
|---|---|---|
| Fall 25 | 56 | 24 |
| Spring 25 | 45 | 21 |
| Fall 24 | 1 | 7 |
| Spring 24 | 1 | 11 |
- 12 contributors, 27 commits, for one lab update
- Workflow was improved by AI
- Content edit aided by AI
II:
Scalability and Efficiency
Registrar Registration Data
Used to do data analysis manually
Student Course Materials Interaction Log
Internal Who-Has-Felt-What feedback
TA asks, bot ships quickly
Exam Stats
Exam Source Files
AI helps analyze, diagnose, and format
Style guide
Drafting policy clarifications
Some TAs don't like having drafting done for them.
Tailored to their taste.
Policy clarifications for staff too
Operations wiki
After deploying all these, one observation:
- Large variance in how quickly staff pick up AI skills
- I have my own setup, but can't replicate it for everyone
- This becomes yet another layer of staff training
Work in progress:
Anigans: my (very virtuous) horcrux
private, secure distillation of my knowledge
🧑💻
👩💻
👨💻
🔁
🔁
engineering-council-talk
By Shen Shen
