Data Science
Data Science
From raw data to decisions
Overview
What this course is about
In 20 weeks (5 months), students go from their first Python script to answering real business questions with data: cleaning messy datasets, writing SQL, running statistical tests and A/B tests, building dashboards, predicting and forecasting with ML, and presenting recommendations to decision-makers. Every project ends with a short written takeaway, because a data scientist's job is to help someone decide.
- Starts from zero: no prior coding or statistics required
- 19 portfolio projects plus a team capstone with demo video
- Grading is 90% hands-on work, marked on analysis and communication
- Runs on student laptops and free tiers; no paid software
Who it's for
Undergraduate students from any branch. No prior coding or statistics required; Class 12 maths is enough.
Format
Two 90-minute theory + live-coding sessions and one 3-hour lab each week (about 6 contact hours per week, ~120 hours total), plus 3-4 hours of self-study and project work.
What you need
Any laptop with 8 GB RAM. Google Colab and Kaggle notebooks work as fallbacks.
You walk away with
A portfolio of 19 projects on GitHub / Kaggle, a public dashboard, and one capstone with a demo video.
Learning outcomes
By the end, you will be able to
- LO1Write clean Python and use Pandas and NumPy to clean, reshape and combine real-world data from files, APIs and websites.
- LO2Query and design relational databases with SQL, including joins, CTEs and window functions.
- LO3Visualise data clearly and build interactive dashboards for business users.
- LO4Apply descriptive and inferential statistics, hypothesis tests, A/B tests and regression analysis correctly.
- LO5Build, evaluate and explain ML models for prediction, segmentation, recommendation and time-series forecasting.
- LO6Analyse text data (including with LLMs) and process large datasets with PySpark.
- LO7Deploy a model or dashboard, check it for bias and privacy risks, and communicate findings to non-technical audiences.
Syllabus
20-week roadmap
Four phases, each ending in a phase gate you prove with a project. Open any week to see topics, the lab and the project you'll ship.
Python, data wrangling and SQL
Most of a data scientist's time goes into getting and cleaning data. This phase builds fluent Python and Pandas, data collection from APIs and websites, and two weeks of SQL, the tool used in almost every data job and interview.
Load raw data into a database, clean it, and answer business questions with SQL and Pandas.
01Python basicsCricket scorecard analyser+
Topics
- Course intro: what data scientists do, with real case studies
- Installing Python 3.12+, Jupyter, VS Code; Google Colab
- Variables, data types, strings, f-strings
- Input and output
- Conditionals and loops
- Functions: parameters, return values, scope
- Git and GitHub basics
- Using AI coding assistants responsibly
You will
- Set up Python, Jupyter, Colab and Git.
- Write programs with variables, conditionals, loops and functions.
- Use AI coding assistants while understanding every line.
Lab · 3 hours
Pair-program a number-guessing game, then refactor it into functions. Push to GitHub.
Project: Cricket scorecard analyser
Reads ball-by-ball data for an innings and reports runs, strike rates, boundaries and top performers.
- • At least 4 functions with clear names
- • Handles bad input lines without crashing
- • Summary printed as a neat table
- • Repo with README
Stretch: Add a partnership analysis.
02Data structures, files, OOP and GitStudent records manager+
Topics
- Lists, tuples, dictionaries, sets
- Comprehensions
- Reading / writing CSV and JSON
- Exceptions and error handling
- Classes, objects, methods; dataclasses
- Modules and packages; virtual environments
- Project structure
- Git branches and pull requests
You will
- Choose the right data structure for a problem.
- Read and write CSV and JSON files.
- Model problems with classes and dataclasses.
Lab · 3 hours
Build a contact book with add / search / delete saved to JSON.
Project: Student records manager
Add, search, update and summarise student records stored in CSV.
- • Student modelled as a class or dataclass
- • Data persists between runs
- • Summaries by department and year
- • Input validation
Stretch: Export a summary report to JSON.
03NumPy and Pandas fundamentalsMovies / OTT titles analysis+
Topics
- NumPy arrays, broadcasting, vectorised maths
- Pandas Series and DataFrames; loading CSV / Excel
- Selecting, filtering, sorting, new columns
- groupby and aggregations
- merge / join, concat
- pivot_table, apply, string and date methods
You will
- Use NumPy arrays and vectorised operations.
- Load, select, filter and transform data with Pandas.
- Aggregate and combine tables.
Lab · 3 hours
Answer 10 guided questions on a sales dataset using only Pandas.
Project: Movies / OTT titles analysis
Answer 15 business questions about a streaming catalogue with Pandas.
- • Questions cover filtering, grouping, merging and dates
- • Each answer shown as a table or number
- • Takeaway paragraph for a content team
- • Notebook runs top to bottom
Stretch: Add 3 charts (preview of Week 7).
04Data cleaning and data collectionMessy-data clean-up+
Topics
- Missing values, duplicates, wrong types, inconsistent text, outliers
- Dates and times
- Tidy data; melt and pivot; combining many files
- Public APIs with requests; JSON to DataFrame
- Web scraping with BeautifulSoup; robots.txt and ethics
- Excel / Google Sheets for analysts: pivots, lookups; Excel vs Pandas
You will
- Clean messy real-world data systematically.
- Collect data from APIs and websites ethically.
- Move analyst work from Excel to Python.
Lab · 3 hours
Scrape a permitted public table (e.g. a Wikipedia list) and clean it into a DataFrame.
Project: Messy-data clean-up
Clean a deliberately messy real dataset (e.g. Indian real-estate listings) and publish a tidy version.
- • Every cleaning step documented with reasons
- • Before / after quality checks (nulls, types, duplicates)
- • Data dictionary for every column
- • Clean file saved as CSV and Parquet
Stretch: Write reusable cleaning functions with tests.
05SQL fundamentalsE-commerce query set+
Topics
- Relational databases, primary and foreign keys
- SQLite and PostgreSQL setup; DBeaver
- SELECT, WHERE, ORDER BY, LIMIT, DISTINCT
- CASE, NULL handling, LIKE, IN, BETWEEN
- Aggregates: COUNT, SUM, AVG, MIN, MAX
- GROUP BY and HAVING
- INNER, LEFT and self joins
- Reading query results critically
You will
- Explain tables, keys and relationships.
- Filter, sort, aggregate and join data with SQL.
- Translate business questions into queries.
Lab · 3 hours
SQL drill: 20 guided queries on a college database, checked against expected results.
Project: E-commerce query set
Write 25 queries of increasing difficulty on an orders / customers / products database.
- • Covers filters, aggregates, GROUP BY / HAVING and 5+ joins
- • Each query answers a stated business question
- • One-line explanation and output for each
- • Saved as a .sql file
Stretch: Solve 20 extra problems on an online SQL practice platform.
06Advanced SQL and SQL with PythonRetail sales analysis (Phase 1 project)+
Topics
- Subqueries and CTEs (WITH)
- Window functions: ROW_NUMBER, RANK, LAG, running totals
- Views; ER diagrams, normalisation, indexes
- sqlite3, SQLAlchemy, pandas.read_sql
- When to use SQL vs Pandas
- Analyst patterns: cohorts, retention, funnels, month-over-month growth
You will
- Use subqueries, CTEs and window functions.
- Design a simple normalised database.
- Combine SQL and Pandas in one workflow.
Lab · 3 hours
Timed practical SQL test (graded), then build a funnel query together.
Project: Retail sales analysis
Load raw CSVs into PostgreSQL and answer business questions with SQL and Pandas.
- • Schema designed and loaded from raw CSVs
- • 10 business questions answered, including a cohort retention table
- • At least 3 queries use CTEs or window functions
- • One-page memo with 3 recommendations
- • 5-minute presentation in lab
Stretch: Automate the load with a Python script.
Visualisation, statistics and analytics
Students learn to explore data visually, tell real patterns from noise with statistics, design and analyse A/B tests, and present findings in dashboards and memos that decision-makers actually use.
Test a claim or experiment correctly and present results in a dashboard and a one-page memo.
07Visualisation and EDAChart makeover + interactive EDA+
Topics
- Principles: chart choice, colour, labels, clutter
- Misleading charts: truncated axes, dual axes, 3D
- Matplotlib and Seaborn
- Interactive charts with Plotly
- The EDA loop: questions -> charts -> insights
- Maps and geo-visualisation basics
You will
- Choose the right chart for a question.
- Spot and fix misleading charts.
- Run a structured exploratory analysis.
Lab · 3 hours
Critique 10 real charts from the news in small groups.
Project: Chart makeover + interactive EDA
Redesign 5 bad charts, then build an interactive EDA of an Indian open dataset (e.g. state-wise air quality).
- • Before / after for each makeover with reasons
- • At least 8 EDA charts, 3 interactive
- • 10 written insights
- • Notebook published on Kaggle or GitHub
Stretch: Add a choropleth map of India.
08Descriptive statistics and probabilityIPL player consistency report+
Topics
- Mean, median, mode; spread, IQR, percentiles, skew
- Box plots and outliers
- Probability rules; conditional probability
- Bayes' rule with real examples (medical tests, spam)
- Binomial, Poisson, normal distributions; z-scores
- Simulation with NumPy
You will
- Summarise data with the right statistics.
- Reason with probability and conditional probability.
- Recognise common distributions.
Lab · 3 hours
Simulate the birthday problem and a medical-test false-positive scenario.
Project: IPL player consistency report
Rank batters and bowlers on consistency, not just averages, using descriptive statistics.
- • At least 4 statistics used and justified
- • Distributions visualised per player
- • Consistency ranking with a clear method
- • Takeaway for a team selector
Stretch: Estimate win probability for common match situations.
09Inferential statisticsTest five claims with data+
Topics
- Populations, samples, sampling distributions
- Central limit theorem by simulation
- Confidence intervals; bootstrapping
- Hypothesis testing: null, alternative, p-values
- t-tests, chi-square, ANOVA
- Errors, power, statistical vs practical significance, p-hacking
You will
- Explain sampling and the central limit theorem.
- Build and interpret confidence intervals.
- Run and interpret common hypothesis tests.
Lab · 3 hours
Bootstrap a confidence interval and compare with the formula.
Project: Test five claims with data
Test real claims (e.g. 'students with coaching score higher', 'weekend sales are higher').
- • Correct test chosen and assumptions checked
- • Effect sizes and confidence intervals reported
- • Plain-English verdict for each claim
- • Limitations stated
Stretch: Apply a multiple-testing correction and discuss the change.
10A/B testing and regression analysisA/B test and pricing analysis+
Topics
- Experiment design: metrics, randomisation, sample size, duration
- Analysing A/B tests
- Pitfalls: peeking, novelty effect, Simpson's paradox
- Correlation vs causation; confounders
- Linear and multiple regression with statsmodels
- Interpreting coefficients; dummy variables; checking residuals
You will
- Design and analyse an A/B test.
- Separate correlation from causation.
- Fit and interpret regression models.
Lab · 3 hours
Calculate required sample sizes for three experiment scenarios.
Project: A/B test and pricing analysis
Analyse a website A/B test and estimate how price and discounts affect sales.
- • Ship / don't ship recommendation with evidence
- • Sample-size and power check
- • Regression with interpreted coefficients
- • Memo written for a product manager
Stretch: Check the A/B result separately for mobile and desktop users.
11Dashboards, BI and data storytellingBusiness dashboard + memo (Phase 2 project)+
Topics
- KPI and metric design
- Dashboard layout principles
- Power BI or Tableau Public: data model, visuals, filters
- Streamlit dashboards in Python
- Data storytelling: structure of a memo and a presentation
- Presenting uncertainty honestly
You will
- Design useful KPIs and dashboards.
- Build a dashboard in a BI tool or Streamlit.
- Present findings to non-technical audiences.
Lab · 3 hours
Build a 1-page KPI dashboard from the Phase 1 retail database.
Project: Business dashboard + memo
Build an interactive KPI dashboard for a business dataset (e.g. food delivery or retail chain) and a memo to the 'CEO'.
- • 5-8 well-chosen KPIs with filters
- • Published dashboard link
- • One-page memo with 3 recommendations
- • At least one statistical test supports a recommendation
- • 5-minute presentation
Stretch: Add a 'what-if' control (e.g. discount level).
Machine learning and forecasting
Students build, validate and explain the models businesses ask for most: prices, churn and risk, customer segments, recommendations and demand forecasts, always compared against a simple baseline.
Build, validate and explain a predictive or forecasting model that beats a sensible baseline.
12ML fundamentals and regressionUsed-car price predictor+
Topics
- Supervised vs unsupervised learning
- Train / validation / test splits; overfitting; bias-variance
- Linear regression with gradient descent from scratch (once)
- scikit-learn API: fit, predict, score
- Metrics: MAE, RMSE, R²; baselines
- Ridge and Lasso
You will
- Explain how supervised learning works and generalises.
- Train and evaluate regression models.
- Always compare against a baseline.
Lab · 3 hours
Fit models of increasing complexity and plot training vs validation error.
Project: Used-car price predictor
Predict used-car prices (e.g. CarDekho data) and explain what drives price.
- • Baseline model and at least 2 better models
- • Proper train / test split, no leakage
- • RMSE and R² reported
- • Feature effects explained for a car dealer
Stretch: Build a simple price-check web form.
13Classification and metricsCustomer churn predictor+
Topics
- Logistic regression
- k-nearest neighbours and decision trees
- Confusion matrix, precision, recall, F1
- ROC-AUC and precision-recall curves
- Class imbalance: class weights, resampling
- Choosing thresholds from costs and benefits
You will
- Train and compare classifiers.
- Choose metrics and thresholds using business costs.
- Handle imbalanced classes.
Lab · 3 hours
Build a cost-based threshold chooser for a loan example.
Project: Customer churn predictor
Predict telecom churn and recommend which customers to target for retention.
- • 3+ models compared
- • Threshold chosen using stated business costs
- • Top churn drivers explained
- • Targeting recommendation with expected savings
Stretch: Estimate the profit of the retention campaign.
14Ensembles, feature engineering and explainabilityIn-class Kaggle-style competition (teams)+
Topics
- Random forests
- Gradient boosting: XGBoost, LightGBM
- Feature engineering: encoding, dates, aggregates
- Pipelines, cross-validation, Optuna tuning
- Feature importance and SHAP
- Data leakage: spotting and preventing it
You will
- Use random forests and gradient boosting.
- Engineer features and build pipelines.
- Explain models and avoid leakage.
Lab · 3 hours
Competition kick-off: explore the data and submit a baseline.
Project: In-class Kaggle-style competition
Teams compete on a private leaderboard (e.g. credit default).
- • At least 5 leaderboard submissions
- • Pipeline with cross-validation
- • SHAP explanation of the final model
- • Top teams present their approach
Stretch: Try model stacking.
15Unsupervised learning and recommendationsSegmentation + product recommender+
Topics
- k-means, hierarchical, DBSCAN; choosing k
- RFM analysis (recency, frequency, monetary)
- PCA for visualisation and compression
- Recommenders: popularity, content-based, collaborative filtering
- Market-basket analysis (association rules)
- Evaluating recommendations
You will
- Segment customers with clustering.
- Reduce dimensions with PCA.
- Build simple recommenders.
Lab · 3 hours
Market-basket analysis on grocery transactions.
Project: Segmentation + product recommender
Segment e-commerce customers and build a 'customers also bought' recommender.
- • RFM features + clustering with profiles
- • PCA visualisation of segments
- • Recommender with a simple evaluation
- • Marketing plan per segment
Stretch: Add a cold-start strategy for new users.
16Time-series analysis and forecastingRetail demand forecast (Phase 3 project)+
Topics
- Trend, seasonality, stationarity; decomposition
- Moving averages, exponential smoothing
- ARIMA / SARIMA
- Prophet; ML forecasting with lag features
- Time-based validation (no shuffling)
- Forecast accuracy: MAE, MAPE; prediction intervals
You will
- Decompose time series into trend and seasonality.
- Forecast with statistical and ML methods.
- Validate forecasts correctly over time.
Lab · 3 hours
Forecast a festival-season sales spike and discuss what the model misses.
Project: Retail demand forecast
Forecast store or product demand for the next 8 weeks and recommend inventory levels.
- • Naive baseline + 3 methods compared
- • Time-based validation
- • Prediction intervals shown
- • Inventory recommendation with risks
- • Capstone proposal submitted (1 page)
Stretch: Add holiday and weather effects.
Specialisation, production and capstone
Students analyse text (including with LLMs), process big data with PySpark, deploy models responsibly, and prepare for jobs while finishing the capstone.
Handle text and big data, deploy their work, and deliver a complete capstone.
17Text analytics, NLP and LLMs for data workReview sentiment and topic analysis+
Topics
- Text cleaning, tokenisation, TF-IDF
- Sentiment analysis
- Topic modelling
- Sentence embeddings for search and clustering
- LLM APIs for labelling, extraction, summarisation
- Measuring LLM label accuracy against human labels
You will
- Turn text into features and insights.
- Use embeddings and LLMs to label and analyse text.
- Check AI-generated labels and code for accuracy.
Lab · 3 hours
Label 100 reviews by hand and compare with an LLM's labels.
Project: Review sentiment and topic analysis
Analyse thousands of product or app reviews for sentiment trends and complaint topics.
- • Sentiment trend over time
- • Top complaint topics with examples
- • LLM vs model labels compared on a human-labelled sample
- • Recommendations for the product team
Stretch: Build a searchable review explorer with embeddings.
18Big data and data engineering basicsPySpark pipeline+
Topics
- Chunking, Parquet, DuckDB
- Why distributed computing; Spark architecture
- PySpark DataFrames
- Spark SQL; joins and aggregations at scale
- Cloud warehouses (BigQuery sandbox)
- ETL / ELT, dbt and Airflow (overview)
You will
- Process data larger than memory.
- Use PySpark and Spark SQL.
- Explain the modern data stack.
Lab · 3 hours
Compare Pandas, DuckDB and PySpark on the same large file.
Project: PySpark pipeline
Clean and aggregate a multi-million-row dataset (e.g. NYC taxi trips or Indian railways data).
- • Pipeline reads raw data and writes summary tables to Parquet
- • At least 5 transformations and 3 aggregations
- • Runtime and data sizes reported
- • Short insights from the summary tables
Stretch: Schedule the pipeline to run daily.
19Deployment, monitoring, ethics and privacyDeployed model with model card+
Topics
- Saving models; FastAPI prediction endpoints
- Streamlit apps; Hugging Face Spaces
- MLflow experiment tracking
- Monitoring and data drift
- Bias and fairness checks; explainability
- Privacy and India's DPDP Act basics; anonymisation
You will
- Deploy a model as an API or app.
- Track experiments and monitor for drift.
- Check work for bias and privacy risks.
Lab · 3 hours
Run a fairness check on the Week 13 churn model across customer groups.
Project: Deployed model with model card
Deploy a Phase 3 model as an API or app for others to use.
- • Public API or app with input validation
- • Logging of requests and predictions
- • Fairness check across at least one group
- • Model card with limits and intended use
Stretch: Add a drift report comparing new data with training data.
20Portfolio, careers and demo day+
Topics
- Portfolio: GitHub, Kaggle, blog posts, LinkedIn
- Roles: data analyst, data scientist, ML engineer, analytics engineer
- Interview prep: SQL tests, statistics questions, case studies, take-homes
- Demo day
You will
- Package work into a job-ready portfolio.
- Prepare for data science interviews.
- Present a capstone to a panel.
Lab · 3 hours
Final dry runs, then demo day.
Capstone
Your team capstone
Teams of 2-4 take a real business or social question from raw data to a recommendation backed by analysis, a model and a dashboard, worth 35% of the final grade. Minimum requirements: real data from at least two sources stored and queried with SQL; a statistical analysis (hypothesis test, A/B test or regression) and a predictive or forecasting model that beats a baseline; an interactive dashboard or app; a written recommendation for a named decision-maker with limitations stated; and a GitHub repo with README, model card, fairness and privacy notes and a demo video. Teams may propose their own idea if it meets the minimum requirements.
UPI / card fraud detection
Which transactions should be flagged for review? (Imbalanced classification, thresholds, SHAP)
Crop yield and price forecasting
What will yield and mandi prices look like next season? (Time series, weather APIs, dashboards)
College placement analytics
What predicts placements, and what should the college change? (SQL, statistics, classification)
Air-quality forecasting
When will AQI cross unsafe levels in a city? (Time series, live data, alerts)
E-commerce personalisation
Which products should each customer segment see? (Clustering, recommenders, A/B design)
Real-estate price analytics
Is this flat fairly priced for its area? (Scraping, regression, maps)
Public transport ridership
Where and when should a city add buses or metro trips? (Big data, PySpark, forecasting)
IPL match analytics
Which strategies and player picks improve win chances? (Statistics, ML, storytelling)
Milestones
- Week 16
Proposal: question, stakeholders, data sources and licences, success metric. Instructor approval required.
- Week 17
Data ready: collected and cleaned data in a database, data dictionary, EDA.
- Week 18
Analysis and model: statistical analysis or test, a model that beats a baseline, error analysis.
- Week 19
Deployed: dashboard or app live, model card, fairness and privacy notes.
- Week 20
Demo day: 10-minute presentation + 5-minute Q&A; final report, repo and demo video (up to 5 min).
Assessment
How you're graded
16 non-phase weeks, about 1.6% each
Weeks 6, 11 and 16, 10% each
timed SQL test in Week 6, statistics and ML concept quizzes
proposal 5%, analysis/model/dashboard 15%, final report 5%, demo day 10%
Tools
What you'll work with
Start Data Science at ₹4,999
Send an enquiry and we'll share batch dates and payment details.