Coverage: Software Engineering Intern / SWE Intern, Summer 2027; live technical interviews; India; reports published from July through September 2026. These are interviews conducted during 2026 recruiting for internships that begin in summer 2027.

This topic will continue to expand as more interview experiences come in throughout the rest of the recruiting season.

🍇 The short version

There isn’t one reliable “Google intern question list” here. The clearest signal is simpler: find a sound approach, explain it clearly, and adapt when the problem changes.

1. Expect a wide range—and follow-ups. Reports ran from easy-to-medium to medium-plus-hard, and four of eight candidates saw a problem extended or reframed.
2. Prioritize both graph/tree and array/matrix fundamentals. Four experiences involved BFS, 0-1 BFS, topological sort, tries, LCA, or tree DP; others leaned on prefix sums, pair counting, two pointers, submatrices, and greedy reasoning. DP is possible, even though one offer-holder did not encounter it.
3. Practice the interview, not just the problems. Candidates had to clarify open-ended prompts, explain trade-offs, write readable code, state complexity, and test by hand without autocomplete.
4. Prepare for AI-judgment questions. Two independent reports mentioned LLM use and handling bad outputs. The focus was verification and ownership—not using AI during the coding round.
5. Keep the sample’s limits in mind. All eight detailed reports were India-based, so these are useful signals for that pipeline, not a confirmed US or EMEA process map.

A practical preparation formula: make arrays, matrices, prefix sums, BFS/DFS, and two pointers automatic → add topological sort, tries, LCA, and tree DP → practice changing a working solution under follow-ups → run 45-minute mocks in a plain editor → prepare two honest examples of how you use and verify AI.

🪸 8 detailed candidate timelines, loops, and screens

01: Prefix/suffix first, weighted-tree DP second

Report date: September 2026 · Role: SWE Intern · Location: India · Process: Off-campus

Round 1

1. Given a positive array, remove one element.
2. Split what remains into two non-empty contiguous subarrays with equal sums, if possible.
3. The candidate summarized the useful direction as prefix/suffix reasoning.

Round 2

1. Given a rooted weighted tree, cut edges so the root can no longer reach any original leaf.
2. Minimize the total weight of the removed edges.

The two rounds show the range inside one loop: a careful array invariant first, then a compact but non-obvious tree recurrence.

02: One round, then a rejection 12 days later

Report date: July 2026 · Role: SWE Intern · Location: India · Process: Big Code · Result: Rejected after Round 1

Round 1

1. The base task resembled the competitive-programming problem Same Differences.
2. The follow-up put absolute values on both sides of the equality.
3. The candidate later said the rejection arrived after 12 days.

This account is useful because it preserves both the follow-up and the wait. The 12-day timing is one person’s experience, not a promised response window.

03: BFS and 0-1 BFS, then strings and tries

Report date: July 2026 · Role: SWE Intern · Location: India · Process: Off-campus through Big Code · Result: Rejected

Round 1

1. Two graph tasks covered ordinary BFS and 0-1 BFS.
2. The candidate said the round took about 40 minutes and reached optimal solutions.

Round 2

1. The problem moved to strings and tries.
2. The candidate spent the available time on the first task and reached a working but non-optimal solution.
3. The rejection came by phone a few days later.

This is one of the clearest reminders that “study graphs” means recognizing the right path model, not merely memorizing a DFS template.

04: Two questions in one round, one question plus follow-ups in another

Report date: September 2026 · Role: SWE Intern · Location: India · Result: Not disclosed

Round 1

1. One LeetCode-medium-style two-pointer problem.
2. One harder trie-DFS problem.

Round 2

1. A medium-level variation on tree LCA.
2. The overall format varied between two separate problems and one problem with follow-ups.

Interview mechanics

1. The candidate emphasized clarifying the open-ended prompt before coding.
2. There was no autocomplete and no ready-made test suite; the candidate was expected to invent edge cases, dry-run the code, and state time and space complexity.
3. The round ended with questions about using LLMs in daily work.

This is the most operationally detailed report in the set. It describes what interview readiness looks like after you already know the algorithm.

05: An offer, but not a universal recipe

Role: SWE Intern · Location: India · Process: Big Code · Result: Offer

Interview loop

1. The candidate completed two DSA rounds.
2. They were asked how they use AI in projects and what they do when it fails to produce the intended result.
3. Their preparation advice centered on peer mocks, speaking continuously, clean variable names, complete code, and complexity analysis.
4. They heard back roughly 24 hours after Round 2.

The useful part is the behavior they practiced. The roughly one-day result timing is one anecdote, and their “no DP required” comment conflicts with another report in this same cycle.

06: A fast submatrix solution, then a long wait

Report date: July 2026 · Role: SWE Intern · Location: India · Process: Big Code · Status: Advanced to Round 2

Round 1

1. The prompt involved a 2D grid and finding a submatrix that satisfied sum conditions.
2. A follow-up added conditions to rule out invalid submatrices.
3. The candidate said they solved it in about 25 minutes with a full explanation and rated the problem easy-to-medium.
4. No project questions were asked.

After the round

1. The candidate initially posted because 15 days had passed without an update.
2. They later edited the post to say the Round 2 email had arrived.

This one is a useful antidote to two bad assumptions: finishing quickly does not mean the question was trivial, and positive interviewer feedback does not guarantee an immediate recruiter update.

07: Topological ordering with several moving pieces

Report date: July 2026 · Role: SWE Intern · Location: India · Process: Big Code · Result: Not disclosed

Round 1

1. Start by deciding whether a topological ordering exists.
2. Discuss the order when the edges do not fully determine one sequence, including duplicate-edge cases.
3. Handle a follow-up about the minimum number of edges needed to form the requested ordering structure.

The candidate openly said this was not the exact prompt and that their summary was vague. That limitation matters: prepare the underlying ideas—indegree, cycle detection, non-unique orders, and edge constraints—but do not memorize the bullets above as though they were an official question statement.

08: Matrix and prefix sums first, math and optimization second

Report date: July 2026 · Role: SWE Intern · Location: India · Process: Off-campus through Big Code · Status: Awaiting an update after Round 2

Round 1

1. The candidate described the topic mix as matrix, prefix sum, and greedy.

Round 2

1. The reported mix was math, greedy, and optimization.

Interview mechanics

1. The candidate advised taking a little time to find the optimal direction before coding.
2. They specifically warned against going silent midway through the implementation.
3. Their second round was scheduled quickly after the first; the outcome was still pending when they posted.

There is not enough source detail to reconstruct either prompt responsibly. The honest takeaway is the combination of topics and the communication advice, no more than that.

What repeated often enough to prioritize

Tier 1: Practice until automatic

1. Arrays and matrices: prefix/suffix sums, counting with a map, two pointers, and recognizing when precomputation removes repeated work.
2. Graph fundamentals: BFS/DFS, 0-1 BFS, cycle detection, and topological ordering.
3. Writing complete, readable code in a plain shared editor without leaning on autocomplete or a supplied test harness.

Tier 2: Be ready to explain deeply

1. Tree recursion and DP: define what each subtree returns, handle leaves cleanly, and justify why local choices cover every path.
2. Tries and LCA, including the state and invariants behind the implementation.
3. Greedy and optimization arguments: state the objective, explain why a local move is safe, and compare alternatives rather than jumping straight to code.
4. Follow-ups: say exactly what changed, what stayed invariant, and whether the original complexity still holds.

Tier 3: Higher-level differentiators

1. AI judgment backed by a real project example: what you delegated, how you checked it, where it failed, and what you still owned.
2. Calm interviewer collaboration: ask useful questions, notice hints, and change direction without pretending your first idea was perfect.

Final checklist

Algorithms and problem recognition

☐ I can move from an array or matrix condition to prefix sums, counting, or two-pointer reasoning without guessing.

☐ I can choose between BFS, 0-1 BFS, DFS, and topological sorting—and explain the choice before coding.

☐ I can define a tree-DP subproblem precisely and justify its transitions and base cases.

☐ I can explain why a greedy choice is safe instead of relying on intuition alone.

Live coding

☐ I can turn an open-ended prompt into explicit inputs, outputs, constraints, and edge cases.

☐ I can test code by hand when there is no Run button.

☐ I state time and space complexity before the interviewer has to remind me.

☐ I use readable names and small helper functions without overengineering the solution.

☐ I have done at least two 45-minute mocks in a plain document or editor with autocomplete turned off.

Follow-ups and AI fluency

☐ After solving a problem, I can handle one changed constraint without restarting from zero.

☐ I have one honest example of AI helping me and one example where I caught or corrected a bad output.

☐ I can explain what I would never delegate without verification.

Come hangout with us on Experience~ Share whatever is on your mind and have fun together! 🚀

Like always, these sources come from wherever useful interview information can be found on internet.