Data Science Interview Questions and Answers: Complete Guide for Pune Job Seekers

Landing a data science job in Pune requires more than just knowing Python and Pandas. Pune's IT companies — from the BFSI giants in Hinjewadi to the analytics startups in Baner and Kharadi — run structured interview processes that test statistics, SQL, machine learning, coding, and business thinking in the same session.

This guide covers the most frequently asked data science interview questions in Pune, organised by topic, with detailed answers, common follow-up questions, and tips on what Pune interviewers are specifically looking for. Whether you are a fresher preparing for your first data analyst role or an experienced professional targeting a senior data scientist position, this guide covers every stage of the interview process.

If you are currently enrolled in or looking for a data science course in Pune, use this guide alongside your coursework to benchmark your interview readiness.

---

How Pune Data Science Interviews Are Structured

Most Pune companies run a 3 to 4 round interview process for data science roles:

Round 1 — Online Assessment (60 to 90 minutes)

• Python coding questions (2 to 3 problems)
• SQL queries (2 to 3 problems)
• Statistics MCQs
• Sometimes a case study or data interpretation question

Round 2 — Technical Interview 1 (45 to 60 minutes)

• Statistics and probability in depth
• Machine learning concepts and algorithms
• Python and Pandas hands-on questions
• SQL query writing

Round 3 — Technical Interview 2 / Case Study (60 minutes)

• Business problem framing
• End-to-end ML pipeline discussion
• Past project deep-dive
• Domain-specific questions (BFSI, manufacturing, e-commerce)

Round 4 — HR / Managerial Round (30 minutes)

• Communication and stakeholder management
• Salary negotiation
• Team fit and career goals

---

Section 1: Statistics and Probability Questions

Statistics is the most commonly tested topic in Pune data science interviews — and the area where most freshers are underprepared.

---

Q1. What is the difference between mean, median, and mode? When would you use each?

Mean is the arithmetic average — sum of all values divided by the count. Use it when data is normally distributed and has no extreme outliers.

Median is the middle value when data is sorted. Use it when data is skewed or has outliers — for example, income data in Pune where a few high earners distort the average.

Mode is the most frequently occurring value. Use it for categorical data — for example, the most common job role applied for on a Pune job portal.

Pune interview tip: Interviewers often follow up with: If I told you the mean salary at our company is Rs 12 LPA but the median is Rs 7 LPA, what does that tell you? Answer: the distribution is right-skewed — a few very high earners are pulling the mean up.

---

Q2. What is the Central Limit Theorem and why does it matter in data science?

The Central Limit Theorem (CLT) states that the sampling distribution of the sample mean approaches a normal distribution as the sample size increases, regardless of the population's original distribution — provided the sample size is sufficiently large (typically n > 30).

Why it matters in data science:

• It justifies using normal distribution assumptions in hypothesis testing even when the underlying data is not normally distributed
• It underpins confidence intervals and A/B testing
• It explains why many statistical tests work well in practice on real-world data

Pune interview tip: Be ready to explain CLT in plain language — 'even if individual customer transactions are not normally distributed, the average transaction amount across many samples will be.'

---

Q3. What is p-value? What does p < 0.05 mean?

The p-value is the probability of observing results at least as extreme as the current results, assuming the null hypothesis is true.

A p-value < 0.05 means there is less than a 5% probability that the observed result occurred by chance if the null hypothesis were true. We reject the null hypothesis at the 5% significance level.

Common misconception to avoid: p < 0.05 does NOT mean there is a 95% probability that the alternative hypothesis is true. It only means the data is inconsistent with the null hypothesis at the chosen significance level.

Pune interview tip: Bajaj Finserv and HDFC analytics teams frequently ask about A/B testing and p-values in the context of product feature experiments.

---

Q4. What is the difference between Type I and Type II errors?

Type I error (False Positive): Rejecting the null hypothesis when it is actually true. Example: concluding a new loan product increases defaults when it actually does not.

Type II error (False Negative): Failing to reject the null hypothesis when it is actually false. Example: concluding a fraud detection model has no improvement when it actually does.

The trade-off between Type I and Type II errors is controlled by the significance level (alpha). Lowering alpha reduces Type I errors but increases Type II errors.

Pune interview tip: In BFSI interviews, interviewers often ask: 'Which error is more costly in fraud detection?' Answer: Type II (missing actual fraud) is typically more costly, so fraud models are tuned for high recall.

---

Q5. What is correlation vs causation? Give an example.

Correlation means two variables move together — when one increases, the other tends to increase (positive correlation) or decrease (negative correlation). It does not imply one causes the other.

Causation means one variable directly causes a change in another.

Example: Ice cream sales and drowning incidents are positively correlated — both increase in summer. But ice cream does not cause drowning. The confounding variable is hot weather, which causes both.

In data science, establishing causation requires controlled experiments (A/B tests) or causal inference methods — not just correlation analysis.

---

Section 2: Python and Pandas Questions

---

Q6. What is the difference between a list and a NumPy array in Python?

Use NumPy arrays for any numerical computation in data science — they are 10 to 100x faster than Python lists for large datasets.

---

Q7. How do you handle missing values in a Pandas DataFrame?

There are four main strategies:

1. Drop rows or columns:

```python

df.dropna() # drop rows with any NaN

df.dropna(subset=['salary']) # drop rows where salary is NaN

df.dropna(axis=1, thresh=100) # drop columns with fewer than 100 non-null values

```

2. Fill with a constant or statistic:

```python

df['age'].fillna(df['age'].mean()) # mean imputation

df['city'].fillna('Unknown') # constant fill

df['salary'].fillna(df['salary'].median()) # median (better for skewed data)

```

3. Forward or backward fill (for time series):

```python

df.fillna(method='ffill') # carry last known value forward

```

4. Predictive imputation: Use a model (KNN Imputer, IterativeImputer) to predict missing values from other features.

Pune interview tip: Always explain WHY you chose a strategy — mean imputation distorts distributions for skewed data; median is safer. For categorical columns, mode or a dedicated 'Unknown' category is better than dropping rows.

---

Q8. What is the difference between groupby and pivot_table in Pandas?

Both aggregate data, but they serve different purposes:

groupby is for flexible aggregation along one or more keys:

```python

df.groupby('city')['salary'].mean()

df.groupby(['city', 'role'])['salary'].agg(['mean', 'count'])

```

pivot_table reshapes data into a 2D summary table — like an Excel pivot:

```python

df.pivot_table(values='salary', index='city', columns='role', aggfunc='mean')

```

Use groupby for programmatic aggregation in pipelines; use pivot_table when you need a cross-tabulation for reporting or visualisation.

---

Q9. How do you merge two DataFrames in Pandas? What are the join types?

```python

Inner join — only rows with matching keys in both DataFrames

pd.merge(df1, df2, on='customer_id', how='inner')

Left join — all rows from df1, matching rows from df2 (NaN where no match)

pd.merge(df1, df2, on='customer_id', how='left')

Right join — all rows from df2, matching rows from df1

pd.merge(df1, df2, on='customer_id', how='right')

Outer join — all rows from both, NaN where no match

pd.merge(df1, df2, on='customer_id', how='outer')

```

Pune interview tip: Be ready to explain when you would use each join type with a business example — 'I would use a left join to keep all customers even if they have no transaction history.'

---

Section 3: SQL Questions

SQL is tested in almost every data science interview in Pune — even for roles that are primarily Python-focused.

---

Q10. What is the difference between WHERE and HAVING in SQL?

WHERE filters rows before aggregation — it operates on individual row values.

HAVING filters groups after aggregation — it operates on the result of GROUP BY.

```sql

-- WHERE: filter before grouping

SELECT city, COUNT(*) as job_count

FROM jobs

WHERE city = 'Pune'

GROUP BY city;

-- HAVING: filter after grouping

SELECT city, COUNT(*) as job_count

FROM jobs

GROUP BY city

HAVING COUNT(*) > 100;

```

---

Q11. Write a SQL query to find the second highest salary from an employee table.

```sql

-- Method 1: Using subquery

SELECT MAX(salary) as second_highest

FROM employees

WHERE salary < (SELECT MAX(salary) FROM employees);

-- Method 2: Using DENSE_RANK (handles ties correctly)

SELECT salary

FROM (

SELECT salary, DENSE_RANK() OVER (ORDER BY salary DESC) as rnk

FROM employees

) ranked

WHERE rnk = 2;

```

Pune interview tip: Always use DENSE_RANK for nth-highest queries — it handles ties correctly. If two employees have the same highest salary, DENSE_RANK still assigns rank 2 to the next distinct salary.

---

Q12. What are window functions in SQL? Give an example.

Window functions perform calculations across a set of rows related to the current row, without collapsing the result into a single row (unlike GROUP BY).

```sql

-- Running total of sales by date

SELECT

sale_date,

amount,

SUM(amount) OVER (ORDER BY sale_date) as running_total,

AVG(amount) OVER (PARTITION BY city ORDER BY sale_date ROWS BETWEEN 6 PRECEDING AND CURRENT ROW) as rolling_7day_avg

FROM sales;

```

Common window functions: ROW_NUMBER, RANK, DENSE_RANK, LAG, LEAD, SUM, AVG, COUNT — all with OVER().

---

Section 4: Machine Learning Questions

---

Q13. What is the difference between supervised and unsupervised learning?

Supervised learning: The model learns from labelled data — each training example has an input and a known correct output. The model learns to map inputs to outputs.

• Examples: Linear regression, logistic regression, decision trees, random forest, SVM
• Use cases: Churn prediction, fraud detection, price forecasting

Unsupervised learning: The model finds patterns in unlabelled data — there is no correct output to learn from.

• Examples: K-means clustering, DBSCAN, PCA, autoencoders
• Use cases: Customer segmentation, anomaly detection, dimensionality reduction

Pune interview tip: Be ready with a business example for each — 'We used K-means clustering to segment Pune customers by purchase behaviour for targeted marketing.'

---

Q14. What is overfitting? How do you prevent it?

Overfitting occurs when a model learns the training data too well — including its noise and random fluctuations — and performs poorly on new, unseen data. The model has high variance.

Signs of overfitting:

• Training accuracy is very high (95%+) but validation accuracy is much lower (70%)
• The gap between training and validation loss keeps growing during training

Prevention techniques:

1. Regularisation — L1 (Lasso) or L2 (Ridge) penalises large coefficients
2. Cross-validation — use k-fold CV to evaluate on multiple held-out sets
3. Reduce model complexity — fewer features, shallower trees, fewer layers
4. More training data — overfitting is less likely with larger datasets
5. Dropout — for neural networks, randomly deactivate neurons during training
6. Early stopping — stop training when validation loss starts increasing
7. Feature selection — remove irrelevant or redundant features

---

Q15. Explain the bias-variance trade-off.

Bias is the error from incorrect assumptions in the model — a high-bias model underfits the data (too simple to capture the true pattern).

Variance is the error from sensitivity to small fluctuations in the training data — a high-variance model overfits (too complex, memorises noise).

The trade-off: Reducing bias (making the model more complex) tends to increase variance, and vice versa. The goal is to find the sweet spot where total error (bias² + variance + irreducible noise) is minimised.

---

Q16. What is the difference between precision and recall? When do you optimise for each?

Precision = TP / (TP + FP) — of all the positive predictions, how many were actually positive?

Recall = TP / (TP + FN) — of all actual positives, how many did the model correctly identify?

When to optimise for precision: When false positives are costly. Example: email spam detection — you do not want to mark legitimate emails as spam.

When to optimise for recall: When false negatives are costly. Example: cancer screening or fraud detection — missing a real case is worse than a false alarm.

F1 Score = 2 × (Precision × Recall) / (Precision + Recall) — use when you need a balance between both.

Pune interview tip: BFSI companies (Bajaj Finserv, HDFC) almost always ask this in the context of fraud detection or credit risk — always frame your answer with a business context.

---

Q17. What is cross-validation? Why is it better than a single train-test split?

Cross-validation (specifically k-fold CV) splits the data into k equal folds. The model is trained on k-1 folds and tested on the remaining fold — repeated k times, each time with a different test fold. The final performance metric is the average across all k runs.

Why it is better than a single split:

• A single split can give misleading results if the test set happens to be easy or hard
• K-fold CV uses all data for both training and testing (just at different times)
• It gives a more reliable estimate of how the model will perform on truly unseen data
• It reduces the variance of the performance estimate

Typical values: k = 5 or k = 10. For small datasets, use Leave-One-Out CV (k = n).

---

Q18. What is the difference between bagging and boosting?

Both are ensemble methods that combine multiple models, but they work differently:

Bagging (Bootstrap Aggregating):

• Trains multiple models in parallel on random subsets of the training data
• Reduces variance (prevents overfitting)
• Example: Random Forest

Boosting:

• Trains models sequentially — each model focuses on the errors of the previous one
• Reduces bias (corrects underfitting)
• Examples: XGBoost, LightGBM, AdaBoost, Gradient Boosting

Pune interview tip: XGBoost and LightGBM are the most commonly used models in Pune's data science interviews — be ready to explain their hyperparameters (n_estimators, max_depth, learning_rate, subsample).

---

Section 5: Feature Engineering and Data Preprocessing

---

Q19. What is feature engineering? Give three examples.

Feature engineering is the process of using domain knowledge to create new features from raw data that improve model performance.

Example 1 — Date decomposition:

From a single timestamp column, extract: day of week, month, quarter, is_weekend, days_since_last_purchase. Each of these can be a powerful predictor.

Example 2 — Interaction features:

If you have 'loan_amount' and 'annual_income', create 'debt_to_income_ratio' = loan_amount / annual_income. This ratio is often more predictive than either feature alone.

Example 3 — Text features:

From a job title column, extract: seniority level (junior/mid/senior), department (engineering/marketing/finance), is_management (True/False).

---

Q20. What is the difference between normalisation and standardisation?

Normalisation (Min-Max Scaling): Scales values to a fixed range, typically [0, 1].

Formula: x_scaled = (x - x_min) / (x_max - x_min)

Use when: The algorithm requires bounded inputs (neural networks, KNN) or when the distribution is not Gaussian.

Standardisation (Z-score Scaling): Transforms values to have mean = 0 and standard deviation = 1.

Formula: x_scaled = (x - mean) / std_dev

Use when: The algorithm assumes normally distributed data (linear regression, logistic regression, SVM, PCA).

Pune interview tip: Always mention that tree-based models (Random Forest, XGBoost) do NOT require scaling — they are invariant to monotonic transformations of features.

---

Section 6: Business and Case Study Questions

Pune companies — especially in BFSI and IT services — test your ability to translate business problems into data science solutions.

---

Q21. How would you build a customer churn prediction model for a telecom company in Pune?

This is a classic end-to-end ML problem. Structure your answer as follows:

1. Define the problem:

Churn = customer cancels subscription within the next 30 days. Binary classification problem.

2. Identify features:

• Usage patterns: call duration, data usage, SMS count (last 30/60/90 days)
• Account features: contract type, tenure, monthly charge, payment method
• Service features: number of complaints, support calls, plan changes
• Derived features: usage trend (is usage declining?), days since last top-up

3. Handle class imbalance:

Churn is typically 5 to 15% of customers — use SMOTE, class_weight='balanced', or adjust the decision threshold.

4. Model selection:

Start with Logistic Regression (interpretable), then try Random Forest and XGBoost. Use cross-validation to compare.

5. Evaluation metric:

Use Recall (catch as many churners as possible) and AUC-ROC. Avoid accuracy — it is misleading with imbalanced classes.

6. Business output:

Rank customers by churn probability. Target the top 20% with retention offers — do not waste retention budget on customers who were not going to churn anyway.

---

Q22. How would you detect fraudulent transactions for a Pune-based BFSI company?

Problem framing: Binary classification — fraud (1) vs legitimate (0). Extreme class imbalance (fraud is typically 0.1 to 1% of transactions).

Key challenges:

• Severe class imbalance — standard accuracy is useless
• Concept drift — fraud patterns change over time as fraudsters adapt
• Latency — fraud detection must be near real-time for card transactions

Approach:

1. Feature engineering: transaction velocity (how many transactions in last 1 hour?), location anomaly (transaction in Mumbai when customer is in Pune?), amount deviation from customer's historical average
2. Handle imbalance: SMOTE on training data, or use anomaly detection (Isolation Forest, Autoencoder)
3. Model: XGBoost with class_weight, or a two-stage model (anomaly detection + classification)
4. Evaluation: Precision-Recall AUC (not ROC-AUC — misleading with extreme imbalance)
5. Deployment: Real-time scoring via REST API with sub-100ms latency requirement

---

Section 7: Project and Experience Questions

---

Q23. Walk me through a data science project you have worked on.

Structure your answer using the STAR method (Situation, Task, Action, Result):

Situation: Briefly describe the business context — 'I worked on a project for a Pune e-commerce company that was losing 18% of customers after their first purchase.'

Task: What was your specific role — 'My task was to build a model to identify customers at risk of not making a second purchase within 30 days.'

Action: What did you actually do — data collection, cleaning, feature engineering, model selection, evaluation. Be specific about tools and decisions.

Result: Quantify the outcome — 'The model achieved 82% recall on the test set. The retention campaign targeting the top 15% of at-risk customers reduced first-purchase churn by 23%.'

Pune interview tip: If you are a fresher without industry experience, use your course project — but frame it with a realistic business context. Interviewers at Trimaya TechnoSol-placed companies have confirmed they value well-framed course projects over vague claims of 'industry experience.'

---

Section 8: Pune-Specific Preparation Tips

---

What Pune Companies Test Differently

BFSI companies (Bajaj Finserv, HDFC, Persistent Systems):

• Heavy emphasis on statistics and probability
• Fraud detection and credit risk case studies
• SQL is tested rigorously — window functions, CTEs, complex joins
• Regulatory and explainability awareness (why did the model reject this loan?)

IT services companies (Infosys, Wipro, Cognizant — Pune offices):

• More emphasis on Python coding and Pandas
• Client communication and stakeholder management questions
• Broader ML knowledge (less depth than product companies)
• Certification (AWS, Azure, Google Cloud) is valued

Manufacturing companies (Tata Motors, Cummins, Thermax — Pune):

• Time series analysis and forecasting
• Predictive maintenance use cases
• IoT data and sensor data processing
• Domain knowledge of manufacturing processes is a differentiator

Startups (Baner, Wakad, Hinjewadi):

• End-to-end ownership — from data collection to deployment
• FastAPI or Flask for model serving
• Cloud deployment (AWS SageMaker, GCP Vertex AI)
• Speed and adaptability over deep specialisation

---

Interview Preparation Checklist for Pune Data Science Roles

Statistics (2 weeks):

• [ ] Descriptive statistics (mean, median, mode, variance, std dev, IQR)
• [ ] Probability distributions (normal, binomial, Poisson)
• [ ] Hypothesis testing (t-test, chi-square, ANOVA)
• [ ] Confidence intervals and p-values
• [ ] Correlation and regression

Python and Pandas (2 weeks):

• [ ] NumPy array operations
• [ ] Pandas: groupby, merge, pivot, apply, lambda
• [ ] Data cleaning: missing values, duplicates, outliers
• [ ] Matplotlib and Seaborn: 5 core plot types

SQL (1 week):

• [ ] Joins (inner, left, right, outer)
• [ ] GROUP BY, HAVING, subqueries
• [ ] Window functions (ROW_NUMBER, RANK, LAG, LEAD, SUM OVER)
• [ ] CTEs (WITH clause)

Machine Learning (3 weeks):

• [ ] Supervised: Linear regression, Logistic regression, Decision tree, Random forest, XGBoost
• [ ] Unsupervised: K-means, PCA
• [ ] Evaluation metrics: Accuracy, Precision, Recall, F1, AUC-ROC
• [ ] Overfitting, regularisation, cross-validation
• [ ] Feature engineering and selection

Projects (ongoing):

• [ ] 2 end-to-end projects on GitHub with README
• [ ] At least one project using a real-world dataset
• [ ] Able to explain every decision in the project

For structured preparation with mentorship, mock interviews, and placement support, explore our Data Science course in Pune at Trimaya TechnoSol on FC Road.

---

Frequently Asked Questions (FAQ)

---

Q1. What are the most commonly asked data science interview questions in Pune?

The most commonly asked topics in Pune data science interviews are: statistics and probability (p-value, hypothesis testing, distributions), Python and Pandas (data cleaning, groupby, merge), SQL (joins, window functions, GROUP BY), machine learning concepts (overfitting, bias-variance, precision vs recall), and at least one case study or business problem. BFSI companies in Pune add fraud detection and credit risk scenarios. For structured preparation, our Data Science course in Pune includes dedicated interview preparation modules covering all these topics.

---

Q2. How should a fresher prepare for a data science interview in Pune?

A fresher preparing for data science interviews in Pune should follow this sequence: (1) Build Python and Pandas fundamentals — be able to clean a dataset and run an EDA from scratch. (2) Learn SQL to the level of window functions and CTEs. (3) Study statistics — focus on hypothesis testing, distributions, and regression. (4) Learn 5 to 6 core ML algorithms deeply — not just what they are, but when to use them and how to evaluate them. (5) Build 2 end-to-end projects on GitHub. (6) Practice mock interviews. Our Data Science course in Pune covers all of this with live projects and placement preparation built in.

---

Q3. What is the salary for a data science fresher in Pune?

A data science fresher in Pune (data analyst or junior data scientist) can expect Rs 3.5 LPA to Rs 6 LPA as a starting salary in 2026. Companies in Hinjewadi IT Park and Kharadi typically offer Rs 4 to Rs 6 LPA for freshers with strong Python, SQL, and ML fundamentals. With 2 to 3 years of experience, salaries rise to Rs 8 to Rs 15 LPA. Our placement page has details on the companies and roles our graduates are placed in.

---

Q4. Is SQL important for data science interviews in Pune?

Yes — SQL is tested in almost every data science interview in Pune, including roles that are primarily Python-focused. Most Pune companies include at least 2 to 3 SQL questions in the online assessment round, and interviewers ask SQL questions in technical rounds as well. The minimum SQL level expected is: all join types, GROUP BY with HAVING, subqueries, and at least basic window functions (ROW_NUMBER, RANK, LAG). Our Data Science course in Pune includes a dedicated SQL module covering all these topics.

---

Q5. What Python libraries should I know for a data science interview in Pune?

The core Python libraries tested in Pune data science interviews are: NumPy (array operations, linear algebra basics), Pandas (data cleaning, groupby, merge, pivot), Matplotlib and Seaborn (EDA visualisation), Scikit-learn (model training, evaluation, pipelines), and optionally XGBoost or LightGBM for advanced roles. Jupyter Notebook proficiency is assumed. Deep learning libraries (TensorFlow, PyTorch) are tested only for ML engineer or AI engineer roles — not for standard data analyst or data scientist positions.

---

Q6. How many rounds are there in a data science interview at Pune IT companies?

Most Pune IT companies run 3 to 4 rounds for data science roles: (1) Online assessment — Python coding, SQL, statistics MCQs (60 to 90 minutes). (2) Technical interview 1 — statistics, ML concepts, Python/Pandas hands-on (45 to 60 minutes). (3) Technical interview 2 or case study — end-to-end problem solving, project discussion (60 minutes). (4) HR or managerial round — communication, career goals, salary (30 minutes). BFSI companies like Bajaj Finserv and Persistent Systems sometimes add a domain-specific round focused on BFSI use cases.

---

Q7. What projects should I build to crack a data science interview in Pune?

For Pune's job market, the most impactful projects are: (1) A churn prediction or fraud detection model — directly relevant to BFSI companies in Hinjewadi. (2) A time series forecasting project — relevant to manufacturing companies in Pune. (3) An end-to-end EDA + ML pipeline on a real dataset with a clear business question. All projects should be on GitHub with a clean README explaining the problem, approach, and results. Pune recruiters check GitHub profiles — a well-documented project is worth more than a vague claim of experience. Our Data Science course in Pune includes live projects designed specifically for Pune's job market.

---

Q8. Does Trimaya TechnoSol provide data science interview preparation in Pune?

Yes — Trimaya TechnoSol on FC Road, Pune provides dedicated interview preparation as part of its Data Science programme. This includes mock technical interviews covering statistics, Python, SQL, and ML, resume preparation, GitHub project review, and direct placement support with companies across Hinjewadi, Kharadi, Baner, and Magarpatta. Visit our Data Science course in Pune page for the full curriculum and batch details, or contact us to book a free demo class and speak with a career counsellor.

---

Trimaya TechnoSol is a leading IT training institute on FC Road, Pune, offering data science, Python, data analytics, Java full-stack development, software testing, AWS DevOps, and digital marketing programmes for freshers and career switchers across Pune — including students from Kothrud, Karve Nagar, Warje, Bavdhan, Pashan, Aundh, Baner, Shivajinagar, and Hinjewadi.