Curriculum

Part I — Analytics, classical ML and forecasting

1. Day 1
   Senior Data Scientist Mindset

   Choose a rule, analysis, or model for a concrete business decision and justify the baseline.

2. Day 2
   Problem Framing and Data Modeling

   Define grain, target, horizon, and when each feature is actually available.

3. Day 3
   Probability and Distributions

   Calculate a conditional probability and identify the assumption behind the update.

4. Day 4
   Statistical Inference

   Interpret a confidence interval and distinguish effect size from a p-value.

5. Day 5
   Causal Thinking and Experiments

   Identify a confounder and propose a defensible experiment or qualified observational analysis.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 6
   Exploratory Data Analysis

   Diagnose a consequential quality issue before selecting a model.

2. Day 7
   SQL and Analytical Data Modeling

   Fix an aggregation or join-grain error and explain its effect on a KPI.

3. Day 8
   Visualization and Executive Storytelling

   Choose a chart and write an evidence-based title with one interpretation limit.

4. Day 9
   Feature Engineering and Preprocessing

   Build a train-only preprocessing choice and identify a leakage trap.

5. Day 10
   Analytics Checkpointcheckpoint

   Complete a short cumulative data-quality, SQL, inference, and chart critique.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 11
   Linear and Logistic Regression

   Explain a regression prediction or class probability, the loss, and a relevant assumption.

2. Day 12
   Trees and Ensembles

   Choose a tree ensemble against a baseline and explain its trade-off.

3. Day 13
   Classification in Practice

   Choose a classification threshold using imbalance and false-action costs.

4. Day 14
   Unsupervised Learning

   Judge whether a cluster or anomaly result is useful beyond an attractive plot.

5. Day 15
   Validation and Optimization

   Design a split/tuning process that keeps evaluation evidence independent.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 16
   Metrics and Business Objectives

   Choose an appropriate metric and show one condition where it becomes misleading.

2. Day 17
   Feature Selection and Interpretation

   Interpret feature importance without treating it as an intervention effect.

3. Day 18
   Error Analysis and Debugging

   Use an error slice or residual pattern to prioritize a debugging hypothesis.

4. Day 19
   MLOps Foundations

   Trace a prediction through training, serving, monitoring, and rollback responsibilities.

5. Day 20
   ML Checkpointcheckpoint

   Defend one end-to-end modeling decision using baseline and error evidence.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 21
   Time Series Foundations

   Distinguish trend, seasonality, and breaks in a plotted time series.

2. Day 22
   Autocorrelation and Transformations

   Use an ACF/differencing example without assuming a diagnostic plot proves model suitability.

3. Day 23
   Classical Forecasting

   Compare seasonal-naive and classical forecasts using a valid historical cutoff.

4. Day 24
   ML Forecasting

   Construct one point-in-time-safe lag/rolling feature for a chosen horizon.

5. Day 25
   Time-Aware Validation

   Draw a rolling-origin backtest and identify which observations are available at each cutoff.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 26
   Advanced Forecasting

   Explain how forecast uncertainty and interval assumptions affect an action.

2. Day 27
   Modern Forecasting

   Choose whether a modern forecasting family warrants testing against a strong baseline.

3. Day 28
   End-to-End Forecasting Projectproject

   Interpret and defend the forecast-project result developed in the guided weekend studios.

4. Day 29
   Senior DS/DA/MLE Case Daycase

   Clarify an ambiguous DS/DA/MLE case and defend the next useful step.

5. Day 30
   Part I Master Assessmentassessment

   Complete a cumulative core probe and select evidence-based Part II review priorities.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

Part II — Deep learning, LLMs and generative AI

1. Day 31
   Artificial Neural Networks

   Trace tensor shapes and parameters through a small neural network.

2. Day 32
   Activations and Losses

   Match output/activation and loss to the prediction task.

3. Day 33
   Backpropagation and Optimization

   Calculate or inspect one gradient update and distinguish backpropagation from the optimizer.

4. Day 34
   Training Deep Networks

   Diagnose an overfitting or unstable-training curve and justify one intervention.

5. Day 35
   Deep Learning Labproject

   Explain a small executed neural-network result relative to a classical baseline.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 36
   CNNs

   Calculate a convolution output shape and explain what the learned filter represents.

2. Day 37
   Sequence Models

   Trace sequence state and identify a limitation of recurrent modeling.

3. Day 38
   Embeddings and Representation Learning

   Interpret an embedding similarity result and identify a domain/evaluation caveat.

4. Day 39
   Attention and Transformers

   Trace Query/Key/Value through a small attention example and its dimensional assumptions.

5. Day 40
   Training and Fine-Tuning Deep Modelscheckpoint

   Choose a transfer-learning strategy under data and compute constraints; include cumulative recall.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 41
   NLP Foundations

   Compare a TF-IDF baseline with an embedding approach for a concrete text task.

2. Day 42
   LLM Architecture

   Explain next-token prediction, context limits, and why fluent output may be unsupported.

3. Day 43
   Pretraining and Fine-Tuning

   Choose prompting, retrieval, or fine-tuning based on the missing capability and available evidence.

4. Day 44
   Parameter-Efficient Fine-Tuning

   Explain what LoRA changes and evaluate its memory/data/quality trade-offs.

5. Day 45
   LLM Evaluation and Alignment

   Design a small task-specific LLM evaluation and identify a judging/contamination risk.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 46
   Generative Modeling

   Compare generative model families by objective and practical quality/diversity constraints.

2. Day 47
   Prompt Engineering and Structured Outputs

   Design a structured-output prompt/schema and validate a failing output.

3. Day 48
   Retrieval-Augmented Generation

   Trace a query from retrieval to grounded context and identify the failing stage.

4. Day 49
   Agents and Tool Use

   Choose an agent or deterministic workflow and define tool limits and stopping conditions.

5. Day 50
   GenAI Application Labcheckpoint

   Inspect an evaluated GenAI/retrieval lab result and complete the cumulative core probe.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 51
   RAG Evaluation and Optimization

   Use labelled retrieval and answer-level evidence to choose a RAG experiment.

2. Day 52
   LLMOps

   Trace one LLM request through retries, observability, versioning, and failure recovery.

3. Day 53
   AI Safety and Security

   Recognize an untrusted instruction and propose an enforceable access/tool boundary.

4. Day 54
   AI Cost, Latency and Reliability

   Calculate a transparent cost/latency trade-off without assuming bigger models are better.

5. Day 55
   Production GenAI Casecase

   Defend a production GenAI design against quality, cost, security, and reliability constraints.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

1. Day 56
   Neural Network Projectproject

   Review actual neural-project artefacts and identify what the baseline comparison proves.

2. Day 57
   LLM Projectproject

   Review a document-search/QA evaluation, its sources, and its unsupported-answer behavior.

3. Day 58
   Production AI System Designcase

   Defend one ML/AI architecture decision using the persisted assumption and decision log.

4. Day 59
   Senior Interview Daycase

   Answer and revise a senior interview case; keep the full interview session optional.

5. Day 60
   Final Assessment and Capstoneassessment

   Complete a final core defense and inspect an evidence-based scorecard; offer resumable capstone/hiring review.

6. SatStudio — build and investigate · 90 min
7. SunStudio — consolidate and defend · 90 min

Bonus studios

Optional 90-minute extras. They never count towards the 24 weekend studios or the cursor; use them to deepen a week or to catch up.

1. Week 2Bonus studio 1: causal inference and experiment review - what the data can and cannot say about price · 90 min
2. Week 2Bonus studio 2: SQL cohort and window analytics on the airline sample database · 90 min
3. Week 2Bonus studio 3: data-quality audit and one-page EDA memo on the messy flight table · 90 min
4. Week 4Bonus studio 4: from a model result to a one-page decision memo · 90 min
5. Week 4Bonus studio 5: classical-ML error analysis and debugging on a classifier · 90 min
6. Week 12Bonus studio 6: interview loop dry run - case, system design and behavioural · 90 min