10 Weeks AI Data
Science Course

This phase focuses on the fundamental shift in mindset and mathematical intuition required to move from reporting to modeling.

• Mindset Shift: Distinguishing between dashboards and data science, moving from descriptive reporting to causal thinking and problem framing.
• Industry Reality: Understanding the end-to-end data science lifecycle (problem → signal → model → decision) and the role of experimentation.
• Applied Mathematics: A refresher on linear algebra (vectors, dot products), probability intuition, optimization, and the mechanics of gradient descent.

This phase covers the essential algorithms and rigorous workflows used to build and validate predictive models.

• Supervised Learning: Deep dives into Regression (regularization, multicollinearity) and Classification (threshold tuning, precision-recall tradeoffs).
• Tree-Based Models: Mastering Decision Trees, Random Forests, and Gradient Boosting (XGBoost, LightGBM) while avoiding feature importance pitfalls.
• Evaluation & Engineering: Learning cross-validation, cost-based evaluation, and the most important skill—feature engineering (lag features, encoding, aggregations).
• Reproducibility: Implementing ML pipelines and versioning to prevent train-test leakage and ensure consistent workflows.

This phase expands into specialized domains and complex data types encountered in business environments.

• Unsupervised & Time Series: Using clustering (K-Means, DBSCAN) for behavioral discovery and ARIMA vs. ML-based forecasting for revenue and demand.
• Causal Thinking: Understanding A/B testing pitfalls, selection bias, and the difference between observational and experimental data.
• Specific Use Cases: Building systems for Anomaly and Fraud detection, an introduction to Deep Learning, and NLP techniques like embeddings and topic modeling.

Labeled as the "2026-relevant" section, this phase integrates modern LLM capabilities into the data science workflow.

• LLM Fundamentals: Understanding how LLMs work, their strengths/weaknesses, and handling hallucinations in a DS context.
• Engineering & Search: Mastering prompt engineering (Chain-of-thought, SQL generation) and building semantic search using embeddings and vector search.
• Agentic Workflows: Developing autonomous data analysis agents capable of tool-calling, multi-step reasoning, and Auto-EDA.
• Human-in-the-Loop: Defining where AI should stop and establishing validation strategies for AI-collaborative patterns.

The final phase focuses on putting models into production and communicating their value to the business.

• Operationalization: Learning model deployment basics (batch vs. real-time), monitoring for data drift, and implementing SHAP for model transparency and fairness.
• Business Integration: Turning predictions into actionable decision rules, estimating ROI, and communicating effectively with stakeholders.
• Capstone Project: A comprehensive build phase involving problem framing and GenAI-assisted analysis, concluding with a formal presentation, decision defense, and portfolio positioning.