IIT Patna AI ML qualifier test: Syllabus, pattern and sample questions
If you are applying to the Certification in Artificial Intelligence and Machine Learning from Vishlesan i-Hub, IIT Patna, offered with Masai, one question comes up quickly: what will the qualifier test ask, and how should I prepare?
This guide covers the admission process, a topic-wise syllabus for the test, a practical look at the pattern, and 10 practice questions with answers. You don't need a computer science degree to follow it. The program is open to anyone who is 12th pass or above, and the test is meant to check your aptitude and readiness, not to filter out beginners.
Quick snapshot of the program
How admission works
According to the program page, the process has three steps:
- Submit an application. You fill in a form describing your motivation and goals. The application fee is ₹100 and is non-refundable.
- Complete the selection and enrollment session. Only shortlisted applicants are invited to this stage, which is where the qualifier or screening takes place.
- Start learning. Once selected and enrolled, you begin the program.
Fees at a glance: a registration fee of ₹4,000 and a program fee of ₹50,000, for a total of ₹54,000 plus GST. You can pay upfront (₹59,720 inclusive of GST) or through an NBFC partner's EMI plan of ₹7,631 per month for 9 months. Fees and batches change, so confirm them on the official program page before you pay.
Section 1: The qualifier test syllabus
Because the program is beginner-friendly and builds from Python basics to generative AI, the qualifier is best prepared for as a test of foundations, logic and learning aptitude and not advanced AI knowledge. The topics below map directly to the course modules, so they are the right places to focus.
Section A: Programming and data handling
- Python basics: variables, data types, loops, conditionals, functions
- Lists, dictionaries, sets, tuples and list comprehensions
- Reading small code snippets and predicting the output
- Basic SQL: SELECT, WHERE, GROUP BY, HAVING, JOINs
- Introductory NumPy and Pandas operations
Section B: Mathematics, probability and statistics
- Mean, median, mode, variance and standard deviation
- Basic probability, including independent events and conditional probability
- Percentages, ratios and simple algebra
- Reading charts and interpreting data
Section C: Machine learning fundamentals
- Supervised vs. unsupervised learning
- Regression vs. classification
- Overfitting, underfitting and train/test splits
- Evaluation metrics: accuracy, precision, recall, F1
- Data leakage and cross-validation, at a conceptual level
Section D: AI and generative AI awareness
- What neural networks and deep learning are, at a high level
- The idea behind Transformers and attention
- What LLMs, prompt engineering and RAG are
- Real-world uses of AI and its limits, such as hallucination and bias
Section 2: Pattern and what to expect
Masai shares the exact test format with shortlisted candidates, so treat the table below as the part to verify with the admissions team.
You can still prepare with confidence. For a beginner-friendly program like this, expect questions that test logical reasoning, basic Python and data skills, and comfort with simple numbers more than rote memorisation. If you need clarification, the program page lists a WhatsApp number (+91 87929 74750) and an email (iitp.programs@masaischool.com).
Section 3: How to prepare (a simple 4-week plan)
- Week 1: Python. Practise short programs daily. Focus on loops, functions, lists and dictionaries, and get used to tracing code by hand.
- Week 2: Maths and statistics. Revise averages, probability and percentages. Practise a few data-interpretation questions each day.
- Week 3: SQL and data basics. Learn how SELECT, GROUP BY, HAVING and JOIN work, and try basic Pandas operations on a small dataset.
- Week 4: ML and AI concepts. Read plain-language explanations of overfitting, evaluation metrics, neural networks, Transformers and RAG. Finish by attempting the practice questions below under a timer.
Two habits matter more than any resource. Work every question out on paper before you check the answer, and note down the concepts you got wrong so you can revisit them.
Section 4: 10 sample questions with answers
These are practice questions written for this guide. They are not official test questions, but they follow the topics above.
Section A: Programming and data handling
Q1. What is the output of the following code?
python
print(len({1, 2, 2, 3}))
(a) 4 (b) 3 (c) 2 (d) Error
Answer: (b) 3. A set stores only unique values, so {1, 2, 2, 3} becomes {1, 2, 3}, which has 3 elements.
Q2. What does this code print?
python
nums = [1, 2, 3, 4, 5, 6]
print([x**2 for x in nums if x % 2 == 0])
(a) [1, 9, 25] (b) [4, 16, 36] (c) [2, 4, 6] (d) [1, 4, 9, 16, 25, 36]
Answer: (b) [4, 16, 36]. The condition keeps only even numbers (2, 4, 6), and each one is squared.
Q3. Which SQL clause filters groups after aggregation, for example "customers whose total spend exceeds ₹1,000"? (a) WHERE (b) ORDER BY (c) HAVING (d) LIMIT
Answer: (c) HAVING. WHERE filters individual rows before grouping. HAVING filters the aggregated results, as in GROUP BY customer_id HAVING SUM(amount) > 1000.
Section B: Mathematics, probability and statistics
Q4. For the data set 2, 4, 4, 6, 9, what are the mean and the median?
Answer: Mean = 5, median = 4. The sum is 25 and there are 5 values, so the mean is 25 ÷ 5 = 5. The middle value of the sorted list is 4.
Q5. Two fair six-sided dice are rolled. What is the probability that the sum is 7?
Answer: 1/6. There are 36 equally likely outcomes. Six of them sum to 7: (1,6), (2,5), (3,4), (4,3), (5,2), (6,1). So the probability is 6/36 = 1/6.
Section C: Machine learning fundamentals
Q6. A model scores 99% accuracy on training data but only 70% on unseen test data. What is the most likely problem, and how can you reduce it?
Answer: Overfitting. The model has memorised the training data instead of learning general patterns. You can reduce it with more training data, a simpler model, regularisation, cross-validation, or early stopping.
Q7. A spam classifier produces 40 true positives, 10 false positives and 20 false negatives. Calculate precision and recall.
Answer: Precision = 0.80, recall ≈ 0.67. Precision = TP ÷ (TP + FP) = 40 ÷ 50 = 0.80. Recall = TP ÷ (TP + FN) = 40 ÷ 60 ≈ 0.67.
Q8. You normalise your whole dataset using its overall mean and standard deviation, and then split it into training and test sets. What is wrong with this approach?
Answer: Data leakage. The test set's information influenced the scaling parameters. The correct approach is to split first, fit the scaler on the training data only, and then apply it to the test data. Leak-free pipelines are a core part of reliable ML work.
Section D: AI and generative AI awareness
Q9. Which architecture, built around the "attention" mechanism, powers modern large language models? (a) Decision tree (b) Transformer (c) K-means (d) Naive Bayes
Answer: (b) Transformer. Transformers use attention to weigh how relevant each word in a sequence is to every other word, which lets them handle context far better than earlier sequence models.
Q10. What problem does Retrieval-Augmented Generation (RAG) mainly address?
Answer: It grounds an LLM's answers in real, up-to-date or private documents. A RAG system first retrieves relevant passages from a knowledge source and then passes them to the model as context. This reduces hallucination and lets the model answer questions about information it was never trained on.
What happens after you clear it?
Once you are in, the 8-month curriculum is organised into five modules:
- Core programming and analytical foundations: Python, Git, SQL, NumPy, Pandas, visualisation and probability
- Predictive modelling and ML systems: regression, classification, ensembles, clustering and pipelines
- Deep learning, NLP and emerging trends: neural networks, transfer learning, Transformers, YOLO, GANs and diffusion
- Generative AI, LLM engineering and agentic systems: prompt engineering, RAG, LoRA fine-tuning and LangChain agents
- Production deployment, MLOps and capstone: REST APIs, containerised deployment, monitoring and drift
You also build portfolio projects, such as a CNN-based product-image classifier and a customer-churn predictor deployed as an API.
A few conditions are worth knowing. To earn the certificate, you need at least 65% overall attendance and 30% marks in cumulative evaluations. Placement assistance (resume reviews, career coaching and interview support) requires a 7+ CGPA and 65% attendance. And remember that this is a professional certification from Vishlesan i-Hub, IIT Patna, not an IIT degree.
Frequently asked questions
Do I need coding experience to apply?No. The program is open to anyone who is 12th pass or above, and the curriculum starts from Python basics. Some familiarity with logic and basic maths will make the test easier.
Is the qualifier test difficult?For a beginner-friendly program, it is designed around aptitude and fundamentals. Steady practice on the topics above is usually enough preparation.
How long is the program, and how much time will I need?Eight months, at roughly 10 hours per week.
What if I don't clear it the first time?Ask the admissions team about re-attempt policies and upcoming batches. This varies by batch, so don't rely on assumptions.
Ready to take the next step?
If you are comfortable with the topics above, or you are willing to put in a few focused weeks, the next step is to apply. Check the current batch, eligibility and fees on the Certification in AI and ML page, and you can compare other options on the Vishlesan i-Hub IIT Patna program listing.