In [7]:
# fit a linear model to the housing set
def fit_model(X, y):
model = LinearRegression()
model.fit(X, y)
return model
CodeBench replaces the five different tools most data science and AI programs stitch together — live coding environments, custom course content, assignments, grading, and tutoring — with one platform, built around what a student actually submits.
Correct approach. Edge case on line 4 not handled when X has one feature.
Most programs teach with one tool, assign work in another, grade by hand in a third, and hope students actually study for the exam. CodeBench is the one place where the code, the grading, and the tutoring are the same system.
Each component works on its own. Together, they share one record of what a student wrote, submitted, and understood — so grading and tutoring are grounded in real work, not guesses.
Give every student a live, ready-to-run coding environment in the browser — Python, R, Apache Spark, and more — with nothing to install and no setup tickets for your IT team.
Build interactive course content — readings, runnable examples, embedded checks for understanding — customized to your syllabus instead of a generic textbook.
Create, release, and collect coding assignments on a schedule — with versioning, deadlines, and submission tracking handled for you, across every section.
Automated code review that flags what's right, what's wrong, and why — with an instructor still signing off on every score. Cuts grading time by over 70%.
Generates a short, personalized quiz from the code a student actually submitted — confirming they understand their own work, whether AI helped write it or not.
Lets students generate their own practice problems ahead of an exam, drawn from the same material they'll be tested on — so review time is spent, not wasted.
A customized AI tutor available whenever a student is stuck — individualized, interactive, and accurate, modeled on what a good human tutor does in office hours, minus the wait.
Instructors and TAs still review every score — CodeBench's grading framework just does the first pass, so a stack of submissions becomes a stack of decisions instead of a stack of reading. The time back goes to office hours, not spreadsheets.
The seven components aren't separate products bolted together — they're stages of the same pipeline, each one built on what actually happened at the stage before it.
Active Book content and a live JupyterHub environment.
Release and track submissions with the assignment framework.
AI does the first pass; an instructor approves the score.
A personalized quiz checks the student understood their own submission.
QuizCraft generates practice problems ahead of the exam.
The AI tutor is there whenever, on that same material.
Write your own content, release assignments, and let AI-assisted grading handle the first pass — you still approve every score.
A live coding environment, a personalized quiz on your own work, and an AI tutor that actually understands what you submitted.
Deploy across departments with a consistent environment for Python, R, and Apache Spark, and one record of student progress.
We'll walk through the platform with a real assignment from your syllabus — not a canned demo.