Pathway 04 / Operate
Production MLOps & AI Platform Engineering
Learn how to package, release, monitor, and support the models you build.
Explore MLOps →Career pathway / Analytical practice
Applied Machine Learning & Data Science
Move from messy questions to defensible analysis, tested models, and recommendations people can act on.
- Target roles
- Data scientist, machine learning analyst
- Learner level
- Foundational–intermediate
- Format
- 10 business days, online-first cohort
- Commitment
- 6–8 hours each weekday; live sessions in CT
- Prerequisites
- Basic Python, algebra, and curiosity about evidence
- Certificate
- Certificate of pathway completion
The study plan
Six moves from raw data to judgment.
Days 01–02Python and analytical SQL
Build a reproducible working vocabulary for inspecting and joining real-world data.
- Data structures, notebooks, environments, and reproducibility
- SQL joins, windows, grouping, and query reasoning
- Question framing, source notes, and analysis scope
Deliverable: a labor-market dataset analysis with documented assumptions and reusable queries.
Day 03Data preparation and insight
Make data quality and the shape of the evidence visible before modeling.
- Cleaning, exploratory analysis, and effective visualization
- Leakage prevention, missingness, and feature design
- Choosing comparisons that support a decision
Deliverable: a decision-ready analysis report with charts, caveats, and a recommended next step.
Days 04–05Supervised learning
Learn the baseline-first discipline behind useful prediction work.
- Regression, classification, trees, boosting, and baselines
- Cross-validation, feature pipelines, and metric selection
- Thresholds, calibration, and error analysis
Deliverable: a benchmarked prediction service with a model card and error review.
Days 06–07Unsupervised and representation learning
Find structure without inventing certainty where the data cannot support it.
- Clustering, dimensionality reduction, and stability checks
- Anomaly detection, embeddings, and similarity
- Interpreting segments and communicating uncertainty
Deliverable: a segmentation study with validation notes and an action-oriented interpretation.
Days 08–09Experimentation and communication
Connect metrics to decisions while respecting the limits of an experiment.
- Metrics, uncertainty, A/B tests, and practical power questions
- Causal cautions, confounding, and observational evidence
- Stakeholder storytelling with a clear recommendation
Deliverable: a model recommendation memo that states evidence, uncertainty, and decision impact.
Day 10Portfolio capstone
Scope, train, evaluate, explain, and serve an end-to-end model in a coherent case.
- Project scope, data card, training, and reproducible evaluation
- Explainability, serving choices, and responsible limitations
- Business case writing and portfolio presentation
Capstone: a deployed project with data card, model evaluation, business case, and clear next-step recommendation.
By the end
Analysis people can use.
You will be able to frame a data question, make the dataset trustworthy enough to inspect, compare models against a baseline, and communicate what the evidence does—and does not—say.
- 01
Work reproducibly
Keep notebooks, queries, assumptions, and environments organized so the analysis can be revisited.
- 02
Prevent self-deception
Spot leakage, confounding, unstable segments, and metrics that quietly reward the wrong behavior.
- 03
Choose a useful model
Compare baselines and error patterns before reaching for complexity.
- 04
Explain the recommendation
Translate model results into a decision memo with uncertainty and responsible limitations.
“The work got better when I learned to write the caveat before the chart.”
Mentor profile
Dr. Theo Mensah
Applied data scientist and educator focused on reproducible analysis, model evaluation, and the communication gap between evidence and decisions.
Before you begin
Quick answers
How much math is required?
The pathway uses practical statistics, algebra, and metric reasoning. You do not need a graduate mathematics background, but you should be willing to inspect how a method works and where it can mislead.
What makes the capstone portfolio-ready?
The capstone includes a data card, reproducible work, baseline comparison, evaluation, deployment notes, and a business case. It is a clear artifact, not a promise of a particular result.
Related pathways
Keep the evidence moving.
Pathway 05 / Assure
Responsible AI, Security & Governance
Add fairness, privacy, threat modeling, and assurance to analytical systems.
Explore responsible AI →