Nine months at 10 to 12 hours a week is the realistic route to a first data role in India, in this order: Python and pandas, then SQL, then statistics, then machine learning, then one specialisation you can deploy, then a portfolio and applications. The part most roadmaps leave out is that your first job will probably not carry the title data scientist, and that is not a failure of the plan.
Each stage below has a month range, the skills inside it, one thing you build, and a checkpoint that says whether you are allowed to move on. Skipping the checkpoints is how people arrive at month nine with nine half-finished topics. This is the data science counterpart of our data analyst roadmap, which is a shorter climb to a job that hires more freshers.
The nine months at a glance
| Stage | Months | What you learn | What you build | Checkpoint before moving on |
|---|---|---|---|---|
| 1 | 1 to 2 | Python, then pandas and NumPy properly | A notebook that cleans a genuinely messy CSV end to end | You can load, clean, group and plot without a tutorial open |
| 2 | 3 | SQL, from SELECT to window functions | 40 queries against a database with at least four related tables | You can write a three-table join with a GROUP BY unaided |
| 3 | 4 to 5 | Probability, distributions, sampling, hypothesis testing | An A/B test read-out that ends in a written conclusion | You can state what your result does not prove |
| 4 | 6 to 7 | Machine learning with scikit-learn, and how to evaluate it | A model with an honest evaluation and an error analysis | You can explain the cases your model gets wrong, and why |
| 5 | 8 | One specialisation, plus deployment basics | The model running behind an API someone else can call | Somebody else uses it without you in the room |
| 6 | 9 | Portfolio, resume, applications, interview practice | Three finished projects and 60 applications sent | You have sat a first-round technical interview |
Stage 1, months 1 and 2: Python until pandas is boring
Data science starts with Python, and unlike the analyst route it starts there rather than with Excel, because everything in stages 3 to 5 is written in it.
Cover the language basics quickly, then spend most of these two months inside pandas and NumPy: reading files, filtering, groupby, merge, reshaping, handling nulls, and dates in the three formats Indian business data arrives in. Add matplotlib or seaborn for fast charts. You do not need design patterns, web frameworks, or algorithms practice at this stage.
Build one notebook that takes a real, dirty file, with inconsistent city spellings and blank fields and numbers stored as text, and turns it into something analysable, with your reasoning written between the cells. Our Python for data analysis guide covers the slice that matters here.
The failure mode in these two months is watching rather than typing. If you have not written code that broke and then fixed it, you have not done stage 1.
Stage 2, month 3: SQL, which every interview still tests
On 6figr’s skill shares for Indian data scientist profiles, SQL carries the same weight as deep learning, which surprises people who assume the job is all modelling. It is also the thing a first-round interview will test while your Kaggle notebook sits unopened.
Work in this order: SELECT and WHERE, aggregates and GROUP BY, joins, subqueries and CTEs, then window functions. Joins are where people go quietly wrong, because a query that silently drops half the rows still returns something that looks like an answer. Our explainer on SQL joins covers that trap.
Write forty queries against a database with at least four related tables, starting from “how many orders last month” and ending at “rank customers by spend within each city”. Write them; do not read them. A structured course such as SQL: Beginner to Advanced at ₹33,110 exists for this month, because having someone review your queries beats another playlist.
Fees last checked September 2026; confirm current numbers with the provider before enrolling.
Stage 3, months 4 and 5: statistics, the part that separates the two jobs
This is the stage that makes you a data scientist rather than an analyst with a model library installed. Cover probability, common distributions, sampling and bias, confidence intervals, hypothesis testing, and the arithmetic behind A/B tests. You do not need measure theory. You do need to be able to say, out loud, why a 2 percent lift on 400 users means nothing.
Build an A/B test read-out on a real or simulated experiment, ending in a few written paragraphs that include what you could not conclude and why. Interviewers probe that section hardest, because it is where they find out whether you understand your own analysis or are reciting a template.
The mistake here is starting machine learning early because statistics feels slow. Models built on a shaky grasp of sampling produce confident nonsense, and the interview will find it.
Stage 4, months 6 and 7: machine learning, and how to tell when it is wrong
Now scikit-learn. Linear and logistic regression, decision trees and random forests, gradient boosting, clustering, and cross-validation. Spend as much time on evaluation as on fitting: train and test splits done honestly, the difference between accuracy and precision and recall, class imbalance, and leakage.
Build one model on a problem you can describe in a sentence, then write the error analysis. Which cases does it get wrong, what do those cases have in common, and what would you collect to fix it. That write-up is worth more in an interview than three more algorithms.
If you want this taught rather than assembled from tutorials, Machine Learning using Python at ₹38,110 covers this stage on its own.
Stage 5, month 8: one specialisation, and getting it out of the notebook
Pick one: natural language processing, time series forecasting, or recommendation. One. A portfolio with a shallow project in each of the three reads as three tutorials; one project with depth reads as a person who can finish.
Then deploy it. Wrap the model in a simple API, put it somewhere with a URL, and have a friend call it. This is the most under-done step in Indian data science portfolios and the easiest way to stand out. You do not need Kubernetes. You need a working endpoint and a paragraph explaining what it returns.
Stage 6, month 9: the portfolio and the applications
Three finished projects, written up properly: a one-paragraph problem statement, the data source, what you did, what you found, and what you would do next. Publish them somewhere linkable.
Then apply at volume. Sixty applications is a floor, and a referral beats twenty cold ones, so tell the people you already know what you are looking for. Expect the first ten to teach you what is wrong with your resume rather than to produce interviews.
Run interview practice in parallel: SQL questions daily, one project you can narrate cold, and a plain answer to “what would you do if the data looked wrong”. Our data science interview questions set is the shape these rounds take in India.
The first roles in India, and what they actually ask for
Genuine fresher data scientist seats are rare in India. AmbitionBox publishes no 0 to 1 year band for the role at all, and its youngest band starts at 1 to 3 years of experience across 23,900 reported salaries. That absence is a hiring signal, not a data gap.
So the realistic first titles are data analyst, associate data scientist, and the analyst-tagged levels inside services firms. Those seats hire freshers, test stages 1 to 3 hard and stage 4 lightly, and put you next to the data you need to move up. A large share of India’s working data scientists started there. Our guide on how to become a data analyst covers that entry point, and the analytics and data science comparison covers which one your target postings are really describing.
On pay, the sources disagree sharply and it is worth knowing why before you quote a number at anyone. Payscale puts entry level, under one year, at ₹5,95,255 a year on 307 salaries, the largest fresher-specific sample of any source checked, and early career, one to four years, at ₹10,05,147 on 754 salaries. 6figr lists a fresher data scientist figure of ₹18 L. That is roughly three times Payscale’s, and the gap is a population effect rather than an error: 6figr and levels.fyi draw on self-selected contributors at large tech and product companies, while AmbitionBox, Indeed and Payscale cover a broader market including IT services. Take both numbers into a negotiation and neither into a fantasy. Our data scientist salary in India breakdown goes through the bands by experience and city.
One more thing the level data shows. Inside a single services firm, the data-science-tagged levels pay above the business-analyst-tagged ones at similar tenure: at Accenture, 6figr puts the Data Scientist Analyst level at an average of ₹14 L at three to five years against ₹9 L for the Analyst level at three to four years. Which internal track you land on matters as much as which company you join.
Route A: graduates and final-year students
You have hours, so compress. Twenty or more hours a week turns this nine-month plan into five or six months, and a final-year project can absorb stage 5 entirely if you choose the topic deliberately.
Three things to do differently. Aim at the analyst seat on purpose rather than as a consolation, because it is the door that opens. Take the internship over the third certificate. And start applying in month four rather than month nine, because hiring runs on a calendar that does not care about your syllabus.
Route B: working professionals
You have a domain, which is an advantage a graduate cannot buy. A supply chain executive who forecasts real dispatch data walks into an interview with a project nobody else has. Generic public datasets are the default because they are easy, and they read that way.
Plan for 10 to 12 hours a week and nine to twelve months, not five. Keep weekday evenings for learning and one protected weekend block for project work, because projects need uninterrupted time in a way practice questions do not.
The fastest route in is usually internal. Ask for the analytics work in your current team before you apply anywhere, because your employer already trusts you and does not need to take a hiring risk. That move is open to you and not to a fresher, and people ignore it in favour of a job hunt that takes four times as long.
Do not quit your job to study full time. It shortens the runway and adds pressure to accept the first offer, and it buys fewer usable hours than people expect.
What this roadmap leaves out on purpose
Deep learning research, Spark and the big data stack, cloud certifications, R, and chasing a Kaggle leaderboard position. Every one of them appears on somebody’s beginner roadmap, and none of them gets a first data job in India. Add them when a specific role asks, which will usually be after you are hired. It also leaves out collecting certificates, because four of them signal less than one project you can defend for twenty minutes.
Where a course helps, and where it does not
Self-study fails on sequencing and feedback rather than ability. Nobody tells you that your queries are inefficient, that your model is leaking, or that you have spent six weeks too long on Python.
SkilloVilla’s Data Science & AI track costs ₹1,01,999, currently ₹89,999, with additional scholarships based on candidate profile; contact admissions for payment options. It runs live online with 1:1 mentorship from working professionals, real projects and placement assistance rather than a job guarantee. Alumni earn 4 to 15 LPA, as published by the company, with a median package of 9.5 LPA, and hiring partners include Deloitte, PwC, Uber, Flipkart and Razorpay.
Where it is the wrong purchase: if you want a formal degree, if you are aiming at ML research, or if you have already proved you finish things alone. Live cohorts run on fixed class timings, and that structure is the product.
Frequently asked questions
How long does it take to become a data scientist in India?
Nine months at 10 to 12 hours a week is realistic for someone working full time, and five to six months for a student who can give it twenty or more hours. The variable is consistency and feedback rather than ability. People who restart every few weeks routinely take two years to reach the same point.
Can I become a data scientist without a degree in computer science?
Yes, and plenty do. Indian hiring at startups and mid-size companies tests skills over pedigree, though a non-CS candidate should expect a longer search and needs a portfolio that proves Python, statistics and model evaluation. A quantitative degree of any kind carries you further than the specific label on it.
Should I learn Python or SQL first for data science?
Python first, then SQL, which is the reverse of the analyst sequence. Data science work in stages 3 to 5 is written in Python, so it is the foundation everything else sits on. SQL then takes a single focused month and stays tested at every interview afterwards, which is why it gets its own stage rather than being picked up along the way.
Do I need machine learning to get my first data job in India?
Not for the first job, in most cases. The analyst and associate seats that hire freshers test Python, SQL and statistics hard and treat machine learning as a bonus. Stage 4 matters for the second job and for the title, so do it, but do not delay applying until you have it.
Is a data science roadmap different for working professionals?
The sequence is identical; the pace and the entry route differ. Plan on nine to twelve months rather than five, build projects on data from your own industry, and look for an internal move into an analytics team before running an external search. That internal route is usually the fastest way in and is not open to a fresher.
How many projects do I need for a data science portfolio?
Three finished ones, with one of them deployed somewhere callable. Interviewers pick a project and dig into it for twenty minutes rather than counting how many you listed. A project with a written error analysis and a stated limitation beats four notebooks that each end at a high accuracy score.
Start at the right stage, not at stage one
Most people reading this are not starting from zero, and starting from zero when you do not need to is its own way of losing three months.
SkilloVilla’s Data Science & AI track at ₹1,01,999, currently ₹89,999, runs this arc live with 1:1 mentorship, real projects and placement support. Take the checkpoints above into a counselling call, say which ones you can already clear, and find out where in the nine months you should begin.
