Skip to main content

Data Science Roadmap 2026

Turn data into decisions: statistics, experimentation, machine learning, causal thinking, generative AI and communication.

7 stages · 16 units · Intermediate · about 38 hours · Last reviewed 19 November 2026

What is a data science?

Data science is the practice of turning data into decisions using statistics, programming and machine learning. A data scientist cleans and explores data, builds predictive models, designs experiments such as A/B tests, estimates causal effects, and explains the results clearly to people who must act on them.

Data Science fresher salary in India

Typical fresher offers for this role are about 5 to 10 LPA per year. This is an indicative range, not a guarantee. Pay depends on city, company type and how strong your projects and interview performance are.

Step-by-step data science roadmap

  1. 1

    Foundations

    Programming, SQL and statistics.

    • Programming

      Programming for data science means using Python and SQL as daily tools: pandas and NumPy for tables and arrays, and SQL for pulling and aggregating data from databases. This foundations unit gives you the working fluency that every later step, from cleaning to modelling, depends on.

    • Statistics

      Statistics for data science is the maths of learning from samples. It includes descriptive measures such as mean and variance, inference with confidence intervals, hypothesis tests that give p-values, and regression that relates an outcome to predictors. It underpins model validation, experiments and honest conclusions.

  2. 2

    Data Wrangling and Exploration

    Know your data.

    • Data Preparation

      Data preparation is the work done before modelling: collecting data from files, databases and APIs, cleaning missing values, duplicates and wrong types, and creating features that help a model learn. It usually takes the largest share of a project's time and has a direct effect on model quality.

    • Exploratory Analysis

      Exploratory data analysis, or EDA, is the first look at a dataset using summary statistics and charts to understand its shape, quality and relationships before modelling. It also includes data quality checks and documenting what each field means, so findings and assumptions can be reviewed by others.

  3. 3

    Machine Learning

    Predict and explain.

    • Supervised Learning

      Supervised learning trains models on labelled examples to predict a known target, either a number (regression) or a class (classification). The unit covers linear and logistic regression, tree-based models and gradient boosting, plus validation methods such as cross-validation and metrics like precision, recall and AUC.

    • Unsupervised Learning

      Unsupervised learning finds structure in data without labels. This unit covers clustering such as k-means and hierarchical methods for segmentation, dimensionality reduction with PCA, and anomaly detection that flags unusual records. Analysts use it for customer segments, visualising high-dimensional data and finding fraud or faults.

    • Model Interpretation

      Model interpretation is explaining why a machine learning model makes its predictions and checking that it treats groups fairly. It uses tools such as permutation importance, partial dependence and SHAP values, along with fairness metrics across groups, so models can be trusted, debugged and defended to stakeholders.

  4. 4

    Experimentation and Causal Thinking

    What actually works.

    • A/B Testing

      A/B testing is a randomised experiment that compares two versions of a product change to measure its effect on a metric. This unit covers hypotheses, choosing metrics, calculating sample size and power, randomisation, running the test, and analysing results without common errors such as peeking or multiple comparisons.

    • Causal Inference

      Causal inference is the set of methods for estimating the effect of a cause when you cannot simply run a clean experiment. It covers confounders, directed acyclic graphs, matching and propensity scores, difference-in-differences and instrumental variables, which let analysts go beyond correlation with stated assumptions.

    • Forecasting

      Forecasting is predicting future values of a time series such as sales or traffic. The unit covers trend and seasonality, baseline methods, ARIMA and exponential smoothing, machine learning with lag features, and backtesting with rolling origin splits, since random train-test splits are invalid for ordered data.

  5. 5

    Deep Learning and Generative AI

    Newer tools for analysts.

    • Deep Learning

      Deep learning for data scientists means building neural networks with PyTorch and using text embeddings and NLP for analytics tasks such as classification, search and clustering of documents. It sits in the newer-tools stage and is most useful for unstructured data like text and images rather than small tabular tables.

    • LLMs for Data Work

      LLMs for data work means using large language models responsibly in analytics: drafting SQL and code, summarising text, extracting fields from documents, and building agents that call tools to analyse data. It stresses verifying outputs, protecting private data and measuring accuracy rather than trusting fluent answers.

  6. 6

    Data Products and Communication

    Impact over accuracy.

    • Delivering

      Delivering data science work means getting results into people's hands: dashboards and small apps built with tools like Streamlit, and basic model deployment as an API or a scheduled batch scoring job. This unit covers packaging, monitoring for drift and handing a model to engineers.

    • Communication

      Communication in data science is turning analysis into decisions: telling a clear story with data, managing stakeholders, and framing results with metrics and KPIs the business already uses. A correct model has little impact if the audience does not understand the finding, its uncertainty or what to do next.

  7. 7

    Capstone and Career

    Evidence of judgement.

    • Projects

      Data science projects are end-to-end builds that show judgement: a churn or demand prediction with validation, an A/B test analysis report, a forecasting project with backtesting, and an LLM-assisted analytics assistant. They prove you can frame a problem, model it and communicate the outcome.

    • Career Readiness

      Career readiness for data science means preparing for interviews and building evidence of skill. Interviews usually test SQL, statistics, machine learning concepts, experiment design and case reasoning, while a portfolio of clear, reproducible projects shows how you work with messy data and explain decisions.

Frequently asked questions

Ready to practise for interviews?

Practise data science questions with an AI mock interviewer and get feedback on your answers.

More roadmaps