GATE DA 2027: Dates, Eligibility & Exam Pattern

Registration opens

2 Sep 2026

GOAPS portal

Last date

27 Sep 2026

5 Oct 2026 with late fee

Exam window

6–21 Feb 2027

Paper slots unannounced

Result

19 Mar 2027

Valid for 3 years

Dates follow the official GATE 2027 Important Dates page, checked 2026-09-06, which IIT Madras marks liable to change. The registration schedule has been revised. Always confirm on gate2027.iitm.ac.in before acting on any deadline.

GATE Data Science and Artificial Intelligence — paper code DA — combines probability, linear algebra, single-variable calculus, programming, databases, machine learning and AI. Its syllabus is not interchangeable with CS. Check the paper codes accepted by your target programmes before choosing between them.

This guide covers the GATE exam window, eligibility, fees and DA marking structure. The official facts are separated from our suggested study sequence. The syllabus overview below uses selected excerpts from the current IIT Madras PDF, rather than a generic data-science curriculum.

What changed for GATE 2027

IIT Madras announces revised syllabi and changes to the test-paper list for GATE 2027. Check these official updates before reusing an older preparation plan:

1. 1The syllabus was revised. The official website asks candidates to check the updated syllabi. Compare the current PDF for your paper with any older study list; do not assume that every change applies to every paper.
2. 2Robotics and Automation (RA) is a new test paper. It joins the paper list as a standalone paper.
3. 3Textile Engineering and Fibre Science moved. It is now a sectional paper under Engineering Sciences (XE) rather than a standalone paper.

GATE 2027 exam dates and registration schedule

The schedule below follows the official Important Dates page, checked 2026-09-06, rather than a frozen copy of the initial brochure. The examination is scheduled across three consecutive weekends in February 2027, with forenoon (9:30 AM to 12:30 PM) and afternoon (2:30 PM to 5:30 PM) sessions.

As of the source check on 2026-09-06, the paper-wise and session-wise timetable was not announced. The overall GATE examination window does not establish your paper's date. Follow the official site for a later timetable before making travel or session-specific plans.

GATE 2027 DA exam pattern and marking scheme

Data Science and Artificial Intelligence (paper code DA) is a three-hour computer-based test carrying 100 marks across 65 questions — 10 General Aptitude questions and 55 subject questions. Questions carry either 1 or 2 marks.

Duration

3 hours

Computer-based test

Questions

65

10 aptitude + 55 subject

Total marks

100

1 or 2 marks each

Language

English

All papers

Note the structural difference from the engineering papers: DA has no separate 13-mark Engineering Mathematics section. Mathematics is not a bolt-on here — probability, statistics and linear algebra sit inside the subject syllabus itself and are examined as core content.

The official question paper pattern specifies the following marking rules. The attempt advice and worked example below are our editorial interpretation, not an official strategy recommendation.

MSQ and NAT have no negative marking, so a plausible final answer with time left may be worthwhile. But time and attention have an opportunity cost: spending several minutes guessing can displace a question you know how to solve. Prioritise solvable questions, set a return-later threshold in mocks, and use remaining time for considered attempts. MSQs require the exact correct set, while NAT entry must follow the stated precision instructions.

GATE DA 2027 syllabus overview

The 7 official sections below come from the IIT Madras DA syllabus PDF, checked 2026-09-06. These are selected excerpts from our full transcription, not an exhaustive topic list. Section names and quoted coverage retain the official wording; an ellipsis marks shortened text. General Aptitude is additional.

• Probability and Statistics — Counting (permutation and combinations), probability axioms, Sample space, events, independent events, mutually exclusive events, marginal…
• Linear Algebra — Vector space, subspaces, linear dependence and independence of vectors, matrices, projection matrix, orthogonal matrix, idempotent matrix…
• Calculus and Optimization — Functions of a single variable, limit, continuity and differentiability, Taylor series, maxima and minima, optimization involving a single variable.
• Programming, Data Structures and Algorithms — Programming in Python, basic data structures: stacks, queues, linked lists, trees, hash tables; Search algorithms: linear search and binary search…
• Database Management and Warehousing — ER-model, relational model: relational algebra, tuple calculus, SQL, integrity constraints, normal form, file organization, indexing, data types…
• Machine Learning — Supervised Learning: regression and classification problems, simple linear regression, multiple linear regression, ridge regression, logistic regression…
• AI — Search: informed, uninformed, adversarial; logic, propositional, predicate; reasoning under uncertainty topics — conditional independence representation…

Use the DA practice syllabus tree to organise revision, but use the full official PDF to decide scope. Our practice taxonomy is not a substitute for the revised 2027 document.

How to prepare for GATE DA 2027

Our editorial suggestion is to build the mathematical prerequisites alongside small worked problems, then connect them to the named learning methods. Adapt this order if a diagnostic set shows you already know a foundation topic; it is not an official preparation schedule.

Build foundations, then connect them to models

1. 1Probability and Statistics. Distinguish a joint probability from a conditional one before working on Bayes' theorem or conditional expectation. For each distribution, record its support and the assumptions needed to use it.
2. 2Linear Algebra. Work small matrix examples for rank, projections, eigenvalues and singular value decomposition. Then connect those calculations to principal component analysis.
3. 3Calculus and Optimization. Practise limits, derivatives and single-variable extrema. Do not silently expand the exam scope to an entire multivariable optimisation course.
4. 4Programming, Data Structures and Algorithms. Trace Python and the named searching, sorting and graph algorithms by hand. Programming experience is useful, but check your ability to reason without running the code.
5. 5Machine Learning. Compare the named regression and classification methods on small examples. Separate training error from validation error, then study the listed clustering and dimensionality-reduction methods.
6. 6AI and Database Management and Warehousing. Schedule both explicitly: search and inference are not the same topic as model fitting, and warehouse modelling is not covered by SQL practice alone.
7. 7General Aptitude. Include recurring practice within the plan and monitor the time needed for correct answers.

Test concepts without relying on a library API

For a small classifier example, write the predicted labels, count errors and explain what changes if the training or validation set changes. For PCA, identify the matrix being decomposed and the direction retained. These are suggested learning exercises, not official PYQs; fluency with a software package is not evidence that each syllabus concept is understood.

Use past questions without treating them as a forecast

• Match each past DA question to a line of the current syllabus, then record any assumptions its solution uses.
• Use textbook exercises for listed concepts that your PYQ practice has not covered. Keep those exercises visibly separate from official past questions.
• Do not drop a listed topic because you have not seen it in the papers you solved. Past frequency is not a commitment about the next exam.
• For MSQ and NAT, weigh a considered attempt against its time cost. For MCQs, eliminate only options you can reliably justify as wrong; uniform blind guessing has zero expected marks before time cost.
• Check the rounding instruction on numerical questions and practise the on-screen entry process.

Audit errors by the concept that caused them

A wrong probability answer may come from conditioning on the wrong event, not from arithmetic. Write down that distinction and choose a new example that tests it. The GATE DA syllabus tree can organise the error log; check the full PDF when a practice tag and the official wording differ.

GATE 2026 vs 2027: selected syllabus comparisons

Checked 2026-09-06: we compared the text of the official IIT Guwahati GATE 2026 PDF with the official IIT Madras GATE 2027 PDF linked below. The rows are editorial summaries of selected passages, not verbatim transcriptions or an official change log. PDF layout, spacing and hyphenation differences are normalized; section and PDF-page references identify the compared scope. This is not an exhaustive list of additions or removals, and unchanged components are labelled explicitly.

Worked example

Original illustration: a classifier flag is not a posterior probability

This is an original teaching illustration, not an official PYQ or a predicted question. Its purpose is to practise a method under explicit assumptions, not to forecast topic weightage or exam difficulty.

In a hypothetical population of manufactured parts, 2% are defective. A binary inspection classifier flags 90% of defective parts and 5% of non-defective parts. These are assumed exact population probabilities, not estimates from a fitted model. One part is sampled uniformly from that population and is flagged. What is the probability that it is defective? Let D mean defective and F mean flagged; D and not-D form an exhaustive partition. No repeated tests are involved.

1. 1Translate the statements without reversing the conditioning: P(D) = 0.02, P(not-D) = 0.98, P(F | D) = 0.90, and P(F | not-D) = 0.05. The requested quantity is P(D | F), not P(F | D).
2. 2Compute both routes to a flag: P(F and D) = 0.90 × 0.02 = 0.018; P(F and not-D) = 0.05 × 0.98 = 0.049. Total probability gives P(F) = 0.018 + 0.049 = 0.067.
3. 3Apply Bayes' rule: P(D | F) = P(F | D)P(D) / P(F) = 0.018 / 0.067 = 18/67 ≈ 0.268657. Multiplying by 100 gives approximately 26.87%.
4. 4Check with expected counts in 10000 sampled parts: 200 defective parts yield 180 flags, while 9800 non-defective parts yield 490 flags. Of 670 expected flags, 180 are defective. This is an expected-count illustration, not a claim that every finite sample has these counts; the base rate explains why the posterior is far below 90%.

Answer: P(D | F) = 18/67 ≈ 0.268657, or 26.87%. The total flag probability is 0.067, not 0.90.

Source check: 2026-09-06. The conceptual reading below supports the method; the official syllabus PDF supports topic scope only. Neither source supplies or endorses this original worked example.

Preparation checklist

Use these editorial prompts for a weekly plan and individual study sessions. Adjust the workload to your available time and current understanding. They are not an official study schedule or a score guarantee, and this one example does not cover the whole paper.

• At the start of the week, select one probability gap and one gap from another official DA section. Write a specific target for each, such as deriving a posterior or checking the rank of a matrix, rather than scheduling only 'revision'.
• Before each probability calculation, define events and the conditioning direction, draw the partition and check that the denominator includes every route to the observed event; verify that the final probability lies between zero and one.
• In one linear-algebra session this week, solve a small system and substitute the solution back. For a proposed eigenpair, compute Av and λv explicitly instead of trusting a memorised pattern.
• In the next machine-learning session, write the train/validation/test split before any preprocessing. State which data fit the transform and which data evaluate it, then identify one possible leakage path.
• Once this week, solve a timed mixed set including General Aptitude, one Python or algorithm trace and a statistics problem. Log time, the model assumptions and why each wrong option or calculation failed.
• At the next review session, read section 1.4.3 of the linked text and recompute this posterior with a different prior while keeping the flag rates fixed. Redo missed questions after 48 hours and use the remaining errors to set next week's targets.

Eligibility, fees and number of papers

The brochure sets a single academic bar: a candidate currently studying in the third or higher year of any undergraduate degree programme, or who has already completed a government-approved degree in Engineering, Technology, Architecture, Science, Commerce, Arts or Humanities, is eligible. There is no age limit and no cap on the number of attempts.

• Papers: you may appear in one or two test papers. A second paper must be chosen from the allowed two-paper combinations listed in the brochure — the combinations are not free-form.
• Score validity: a GATE 2027 score stays valid for 3 years from the date the results are announced.
• Mode: computer-based test, in English only, conducted at centres across 300+ cities in India. There are no international centres for GATE 2027.
• Two-paper fees: check the table for your category and registration period. The official extended-period two-paper amount is not obtained by doubling the extended one-paper fee.

Age limit

None

No cap on attempts either

Papers allowed

1 or 2

From the allowed combinations

Score validity

3 years

From the result date

Exam centres

300+ cities

India only, no overseas centres

Registration happens only through the GATE Online Application Processing System (GOAPS), reachable from the official website. Nothing is submitted on paper — IIT Madras explicitly asks candidates not to post printed application forms or documents.

Practising GATE DA previous year questions

Previous year questions show how DA topics were examined in those years; they do not predict the next paper. Try an untimed set to diagnose understanding, then a timed paper to check pacing. For each error, record the concept, the incorrect step and a corrected solution you can reproduce without notes.

Use the GATE DA resource hub to find available released papers and question-backed revision companions, or organise study with the GATE DA topic map. The resource hub checks availability before displaying practice links; it also links official past papers when our archive does not yet cover a stream.

Frequently asked questions

Who is conducting GATE DA 2027?

IIT Madras is the Organizing Institute for GATE 2027. Current dates and notices are published at gate2027.iitm.ac.in; check that source before acting on a registration deadline.

What is the GATE DA 2027 exam date?

The GATE 2027 exam window covers 6, 7, 13, 14, 20 & 21 February 2027. As of 2026-09-06, the specific day and session for DA is not yet announced. Check the official paper-wise timetable when published; dates are liable to change.

Does GATE DA have an Engineering Mathematics section?

DA has no separate 13-mark Engineering Mathematics allocation. General Aptitude carries 15 marks and the subject syllabus 85. Probability and Statistics, Linear Algebra, and Calculus and Optimization are official DA syllabus sections within those subject marks.

Should I choose GATE DA or GATE CS for AI and data science programmes?

Check each target programme's accepted paper codes and admission criteria first. DA focuses on data science and AI content; CS also covers systems and theory. Compare the official syllabi against your background. Taking two papers is allowed only for a combination on the official list, and admission is not guaranteed by qualification.

Is there negative marking in GATE DA 2027?

Only MCQs have negative marking: 1/3 mark for a wrong 1-mark answer and 2/3 mark for a wrong 2-mark answer. MSQ and NAT have no negative marking; MSQs have no partial marking. No penalty does not remove the time needed to attempt a question.

How long is a GATE 2027 score valid?

A GATE 2027 score remains valid for 3 years from the date the results are announced. Results are scheduled for 19 March 2027; dates are liable to change.