Foundation & Job Market Readiness
Expert review guide for learners, instructors, and career switchers. This document defines what “good enough” looks like before you advance stages, and how the curriculum maps to skills employers often screen for in Germany, the US, and remote-first teams. It does not guarantee a hire.
Read this with: START-HERE.md · LEARNING_ROADMAP.md · Career Roadmap Guide
Professor’s verdict (short)
| Area | Rating | Notes |
|---|---|---|
| Breadth | Strong | Tabular ML → DL → GenAI → deploy; 23 projects; system-design side track |
| Foundation depth | Strong | Module 00 math + Python is above typical bootcamp level |
| Hands-on ratio | Mixed | Beginner projects ship code; intermediate/advanced are brief-led |
| Job-market alignment | Good with gaps | SQL and portfolio gates come late in the default stage order |
| Assessment | Self-paced only | No autograded mastery checks; use exit criteria below |
Bottom line: Follow stages and role paths, not folder numbers 00→25 in order. Start projects in parallel after Stage 2. Treat SQL as Stage 1.5 if you target analyst or data scientist roles.
Critical rule: stages ≠ folder numbers
| Mistake | Fix |
|---|---|
| Opening folders 00, 01, 02… in numeric order only | Use stage table in START-HERE |
| Doing deployment (13–14) before SQL (19) on an analyst track | Follow Data Analyst path: 00 → 01 → 19 early |
| Waiting until “Stage 9” to build projects | Start beginner projects after Stage 2 while continuing modules |
| Reading entire 1,900-line guides in one sitting | Use learning objectives + exit criteria; skim advanced files later |
Foundation exit gates (do not skip)
Gate A, after Module 00 (Stage 0)
You are ready for Module 01 when you can without copying from a tutorial:
- Write functions, classes, and file I/O (JSON or CSV) in Python
- Explain Big-O for a simple loop vs nested loop
- Multiply matrices and explain why linear algebra matters for ML
- Compute mean, variance, and interpret a normal distribution
- Explain gradient descent in plain language
- Run Jupyter and install packages in a virtual environment
- Capstone: Complete the movie script generator or NumPy NN tutorial
Job relevance: Every technical interview assumes this baseline. Weak Python here causes attrition in Module 01.
Gate B, after Module 01 (Stage 1)
You are ready for Module 02 (ML) when you can:
- Load a CSV with Pandas, handle missing values, and summarize dtypes
- Build at least three plot types (distribution, relationship, categorical)
- Complete a short EDA narrative: question → data → chart → insight
- Use NumPy for vectorized operations (avoid Python loops on large arrays)
Parallel track (analyst / data scientist): Start Module 19 SQL now, not after GenAI. Most EU and US job postings expect SQL alongside Python.
Job relevance: Junior data roles are won on EDA + SQL + communication, not on neural networks.
Gate C, after Modules 02–05 (Stage 2)
You are ready to start entry-level portfolio work for tabular ML roles when you can:
- Build train/validation/test splits and explain data leakage
- Train regression and classification models with scikit-learn
- Choose metrics (RMSE vs F1 vs PR-AUC) for the business problem
- Tune hyperparameters with cross-validation
- Discuss fairness at a basic level (Module 04)
- Portfolio: Finish at least two beginner projects (e.g. house prices + Titanic)
Do not wait for deep learning or GenAI to start GitHub portfolio work.
Gate D: Production skills target (Stages 7–8 + projects)
Aim at ML Engineer / MLOps interview prep when you can:
- Expose a model via FastAPI or Flask with Docker
- Track experiments (MLflow or W&B) and version a model artifact
- Explain monitoring and drift at a high level
- Portfolio: One intermediate + one advanced project with README, tests, and deploy link
Side track: Complete system-design lessons 00–14 before senior ML engineer interviews.
Role-specific “minimum viable” paths
| Role | Minimum modules | SQL timing | First portfolio milestone |
|---|---|---|---|
| Data Analyst | 00, 01, 19, 20, 21 | After 01 | Streamlit dashboard project |
| Data Scientist | 00–08, 15, 19–21 | After 05 or parallel with 06–07 | 3 beginner + 1 intermediate |
| ML Engineer | 00–10, 13–14, 19–21 | After 01 | Deployed API + MLflow run |
| LLM Engineer | 00–01, 05, 09–10, 12, 25, 13–14, 19 | After 01 | RAG app with eval notes |
Full tables: career_roadmap_guide.md
Known curriculum gaps (and workarounds)
| Gap | Why it matters | Workaround in this repo |
|---|---|---|
| No autograded exams | Hard to prove mastery | Use exit gates above; time-series exercises as template |
| SQL late in default stages | Analyst jobs need it early | Stage 1.5 parallel path to Module 19 |
| Intermediate projects lack starter code | Drop-off for guided learners | Pair each brief with a Kaggle dataset + notebook skeleton you create |
| MLOps mostly conceptual | Interviews ask hands-on | deployment.md + mlops.md labs + Docker tutorial |
| Ethics not required | EU AI Act, US enterprise | Read ethics_in_ml.md before Module 04 fairness section |
| Communication not sequenced | DS hires for storytelling | stakeholder_communication.md after first EDA project |
Recommended weekly rhythm (working professional)
| % | Activity |
|---|---|
| 50% | Code (guides, exercises, projects) |
| 25% | One module lesson (read + reproduce examples) |
| 15% | Portfolio / GitHub (README, screenshots, deploy) |
| 10% | Review (quick-ref sheets, spaced repetition) |
Instructor / self-check questions
Before advancing a stage, ask:
- Can the learner explain the last concept to a non-technical colleague?
- Can they reproduce the core notebook without the guide open?
- Is there a GitHub artifact proving the skill?
- For EU learners: can they discuss data minimization and model limitations?
Next steps
- New learner → GETTING_STARTED.md
- Career switch → career_roadmap_guide.md
- Compressed timeline → QUICK_START.md
- Portfolio checklist → career_portfolio.md
- Interview prep → interview_prep.md