The official scikit-learn certification, proctored by Probabl

The senior bar. Set by the people who maintain scikit-learn.

The Expert Practitioner Certification targets senior data scientists shipping to production. Custom estimators, calibration, MLOps, and debugging teammate models.

120 min proctored JupyterLite included Pass at 70%
Probabl
VERIFIED
03 / Expert SKL-E-2026
Certificate of Scikit-learn
expert practitioner
Level Senior data scientist
Credential type Verifiable credential
BY PROBABL
$499USD / exam fee
120 minexam length
36 + 2 labsquestions
70%passing score
3 yearscredential validity
What we evaluate

Seven competencies of a senior data scientist.

The Expert certification ensures certified professionals have production-level mastery of scikit-learn, including custom estimators, calibration, and team-level diagnostics.

01

Expert-level Machine Learning

In-depth algorithmic knowledge and awareness of emerging trends.

02

Algorithm Development

Custom ML algorithm implementation using the scikit-learn estimator API.

03

Model Deployment

Production expertise and MLOps: serialization, serving, and monitoring.

04

Research and Innovation

Independent research capability and ability to apply novel methods.

05

Strategic Planning

Long-term data science strategy involvement and project scoping.

06

Strategic Vision

Understanding industry trends and positioning ML work within broader business context.

07

Model Diagnostics

Pipeline troubleshooting for team members — reading others’ code, finding failure modes, proposing fixes.

Six topics. The shape of the Expert exam.

Production-grade scikit-learn. Custom estimators, calibration, MLOps, and diagnosing pipelines you didn’t write.

01

Machine Learning Concepts

Supervised and unsupervised, model families, loss functions, splitting criteria, feature selection methods, and calibration vs. ranking power.

  • Supervised and unsupervised learning, model families
  • Loss functions and splitting criteria
  • Feature selection methods
  • Calibration vs. ranking power differentiation
from sklearn.metrics import brier_score_loss, roc_auc_score

brier = brier_score_loss(y, p)  # calibration
auc   = roc_auc_score(y, p)    # ranking power
02

Model Building and Evaluation

Custom estimators, metadata routing, calibration plotting, and post-calibration techniques.

  • Custom estimators following the sklearn API
  • Metadata routing
  • Calibration plotting (reliability diagrams)
  • Post-calibration techniques: isotonic, Platt scaling
from sklearn.calibration import CalibratedClassifierCV

cal = CalibratedClassifierCV(
  estimator=clf,
  method="isotonic",
  cv=5,
).fit(X_tr, y_tr)
03

Interpretation and Communication

Partial dependence plots, permutation importance, pipeline diagnosis, feature selection pitfalls, code comprehension.

  • Partial dependence plots
  • Permutation importance
  • Pipeline diagnosis and feature selection pitfalls
  • Reading and explaining others’ code
from sklearn.inspection import (
  PartialDependenceDisplay, permutation_importance,
)

PartialDependenceDisplay.from_estimator(
    clf, X, features=[0, 1],
)
04

Data Preprocessing

Parquet loading, plot interpretation for model selection, multi-source data combining, feature engineering including lagged features.

  • Loading and joining parquet datasets
  • Plot interpretation for model selection
  • Multi-source data combining
  • Feature engineering including lagged features
import pandas as pd

df["sales_lag_7"] = (
  df.groupby("store")["sales"].shift(7)
)
05

Model Selection and Validation

Hyperparameter tuning with proper scoring rules and calibration-focused metrics.

  • Proper scoring rules for probabilistic outputs
  • Calibration-focused metrics in GridSearchCV
  • Custom scorers with make_scorer
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import make_scorer, brier_score_loss

scorer = make_scorer(
  brier_score_loss,
  greater_is_better=False,
  needs_proba=True,
)
GridSearchCV(pipe, grid, scoring=scorer)
06

Model Deployment

Model serialization with joblib, pickle, or skops — trade-offs and security considerations.

  • Serialization with joblib and pickle
  • Secure serialization with skops
  • Trade-offs between formats
import skops.io as sio

sio.dump(model, "model.skops")
loaded = sio.load(
  "model.skops", trusted=True,
)
Certification ladder

Three levels. You are on the third.

Three certifications, each matching a level and a typical data scientist career path.

Associate Practitioner

Junior data scientist

Fundamental ML, preprocessing, evaluation

Professional

Mid-level

Regularization, ensembles, feature engineering, nested CV

LEVEL 03 YOU ARE HERE

Expert

Senior practitioner

Custom estimators, calibration, MLOps, diagnostics

Train where you will be tested

Get training with Skolar.

The Expert track on Skolar covers custom estimators, calibration, MLOps patterns, and model diagnostics, with notebooks and practice written by the scikit-learn team.

01 Completed
Associate Practitioner
8 lessons · ~24 h
02 Completed
Professional
10 lessons · ~32 h
03 4/12 complete
Expert
12 lessons · ~40 h
Continue ↗
The exam, in brief

Logistics, plain.

Everything you need to plan your sitting, in six lines.

Format Proctored online via Webassessor
Duration 120 minutes, 36 multi + 2 labs
Passing 70%, graded by topic area
Languages English, French coming Q4
Fee $499 USD, one retake included
Validity 3 years, renewable via re-examination
Frequently asked

Questions we get a lot.

Are there prerequisites?
No formal prerequisites. Most candidates have years of production scikit-learn experience. The exam represents a senior-level bar.
What are the labs?
Two hands-on sandboxed labs: the first requires debugging a colleague's pipeline; the second demands building a custom estimator with metadata routing. Code is submitted and graded against held-out tests.
What happens if I do not pass?
One retake is included. Additional retakes are available at a discounted rate. A 30-day cool-down between attempts gives you time to revisit weak topics on Skolar.
Is the credential verifiable?
Yes. Passing candidates receive a credential ID and a public verification page on probabl.ai. Recruiters can confirm validity without contacting you.
Does it expire?
The Expert certification is valid for 3 years. Renewal requires re-taking the Expert exam at a discount. Credential identifiers persist across renewals.
Ready when you are

The credential the maintainers sign.

120 minutes. $499 USD. Multiple-choice plus two hands-on labs, a credential issued by the people who ship scikit-learn.