Meta Software Engineer Onsite Questions: 3 Experiences
- Company
- Meta
- Reports
- 3 interview reports
- Level
- Unknown
- Stages
- Onsite
Meta software engineer onsite questions reported in August 2026 by three candidates: an experienced hire, a new grad and one whose level wasn't given.
At a glance
- Experienced full-time (report 01): Two onsite coding tasks, stock profit with a one-trade follow-up and a mouse-and-cheese reachability search, with no round order given. Passed.
- New grad full-time (report 02): Six onsite coding tasks, from a circular linked list and a course schedule to palindromes, word break, BST range queries and sparse vectors, plus a behavioral talk. Passed.
- Level not reported (report 03): A phone screen with two coding tasks, two onsite coding rounds of two tasks each, then a system design round on AND/OR search over user statuses. Result not known.
- Practise first: Validate a Palindrome After Limited Deletions (named in report 02 and report 03), then Maximum Stock Profit with Transaction Limits (report 01), Minimum Time to Complete Target Courses (report 02) and Design a Boolean Status Search Index (report 03).
Coverage: Three Meta onsite reports, all posted in August 2026. Report 01 is an experienced full-time General Software Engineering candidate who interviewed in January to March 2026. Report 02 is a new-grad full-time General Software Engineering candidate who interviewed in October to December 2025. Report 03 is a Software Engineer whose hiring level was not given.
The three hiring levels are kept apart below, one section each: Experienced full-time (report 01), New grad (report 02) and Hiring level not reported (report 03).
Experienced full-time: General Software Engineering
What came up
The one experienced report had two coding tasks: a stock-profit problem with a follow-up, and a reachability question where the only tools are two true-or-false calls.
Questions we found
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| Maximum Stock Profit with Transaction Limits | 1 | Jan–Mar 2026 |
| Mouse and Cheese Reachability | 1 | Jan–Mar 2026 |
Candidate timeline
01: Experienced engineer, stock profit and a mouse hunting for cheese
Interview date: January to March 2026 · Report date: Aug 24, 2026 · Role: General Software Engineering · Level: Experienced, full-time · Result: Pass
The candidate came in through a recruiter.
Onsite coding
- Stock profit, two versions. You get a list of daily stock prices. First, find the most profit you can make when you may buy and sell as many times as you like. The follow-up allowed just one buy and one sell. Practise both versions in one place with Maximum Stock Profit with Transaction Limits.
- Can the mouse reach the cheese? A mouse starts at (0, 0). You only get two calls:
move(x, y), which returns true or false, andreachCheese(), which also returns true or false. The task is to decide whether the mouse can get to the cheese. The report left open how coordinates work and what a failed move does. Our practice version, Mouse and Cheese Reachability, fixes those rules: one step at a time to a neighboring cell, and a failed move leaves the mouse where it was.
New grad full-time: General Software Engineering
What came up
The one new-grad report covers six coding tasks across linked lists, graphs, strings, tree range queries and sparse vectors. Three of them had follow-ups.
Questions we found
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| Insert into a Sorted Circular List | 1 | Oct–Dec 2025 |
| Minimum Time to Complete Target Courses | 1 | Oct–Dec 2025 |
| Validate a Palindrome After Limited Deletions | 1 | Oct–Dec 2025 |
| Word Break with a Reusable Dictionary | 1 | Oct–Dec 2025 |
| Repeated BST Range Sum and Average | 1 | Oct–Dec 2025 |
| Sparse Vector Dot Product from Tuples | 1 | Oct–Dec 2025 |
Candidate timeline
02: New grad, six coding tasks and a behavioral talk
Interview date: October to December 2025 · Report date: Aug 24, 2026 · Role: General Software Engineering · Level: New grad, full-time · Result: Pass
Onsite coding
- Insert into a circular linked list. Add one new value to a linked list whose last node points back to the first. The report gave no further rules. Our version, Insert into a Sorted Circular List, keeps the list in sorted order and allows duplicates.
- Course schedule with durations. Each course has prerequisites and takes some amount of time. The question was the shortest total time to finish a given list of target courses. The talk covered prerequisite groups that are not connected to each other, and a breadth-first approach that works one level at a time. Practise it as Minimum Time to Complete Target Courses.
- Palindrome after deletions. First: can the string become a palindrome if you delete one character? Then the same question with up to k deletions. This is Validate a Palindrome After Limited Deletions, the same task report 03 got.
- Word break. Given a string and a set of words, decide whether the string can be cut into pieces that are all words from the set. Words can be used more than once. See Word Break with a Reusable Dictionary.
- BST range sum, then average. Find the sum of the values in a binary search tree that fall inside a range. Next, return the average for a range. Last, make many repeated range queries fast by doing work up front and using prefix sums (running totals). All three steps are in Repeated BST Range Sum and Average.
- Sparse vector dot product. Multiply two mostly-zero vectors and add up the products. The follow-ups asked about one dense and one sparse vector, and about vectors stored only as (index, value) pairs for the nonzero entries. The pairs version is Sparse Vector Dot Product from Tuples.
Behavioral
The report also mentions a behavioral conversation, with no question details.
Hiring level not reported: Software Engineer
What came up
This report has two grid problems, a target-sum subarray count, a binary search, a string task with a harder follow-up, a tree sum and a search-index design.
Questions we found
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| Find and Reconstruct a Binary-Matrix Path | 1 | Aug 2026 (report date) |
| Count Subarrays with a Target Sum | 1 | Aug 2026 (report date) |
| Find a Deterministic Local Minimum | 1 | Aug 2026 (report date) |
| Validate a Palindrome After Limited Deletions | 1 | Aug 2026 (report date) |
| Query Whether an Island Size Exists | 1 | Aug 2026 (report date) |
| Sum Root-to-Leaf Numbers | 1 | Aug 2026 (report date) |
System Design
| Question | Reports | Last seen |
|---|---|---|
| Design a Boolean Status Search Index | 1 | Aug 2026 (report date) |
Candidate timeline
03: Software engineer, phone screen through system design
Report date: Aug 16, 2026 · Role: Software Engineer
Phone screen
- Path through a blocked grid. In a grid of open and blocked cells, find a way from the top-left corner to the bottom-right one, and return the path itself, not just yes or no. The approach was breadth-first search: save the cell each step came from, then walk back from the end to rebuild the path. Practise it as Find and Reconstruct a Binary-Matrix Path.
- Subarrays that hit a target sum. Count the runs of consecutive numbers that add up to a target, where numbers can be zero or negative. The approach was prefix sums plus a map that counts how often each running total has appeared. See Count Subarrays with a Target Sum.
Onsite coding, round 1
- Local minimum. Find a position whose value is no bigger than its neighbors, using binary search. The edge cases that came up were an array with a single element and neighbors with equal values. Our version, Find a Deterministic Local Minimum, pins down which answer to return when more than one works.
- Palindrome after deletions. Can the string become a palindrome by deleting one character? The follow-up raised that to at most k deletions, solved by caching results for each pair of left and right positions plus the deletions still left. Practise it with Validate a Palindrome After Limited Deletions.
Onsite coding, round 2
- Island size lookup. In a grid of land and water, measure every island once with depth-first search and keep the sizes in a set. After that, a call like
isSizeExist(size)answers in constant time. This is Query Whether an Island Size Exists. - Root-to-leaf numbers. Each tree node holds a digit, and each path from the root down to a leaf spells a number. Add up all those numbers. The approach was depth-first search, and the talk turned to a loop-based version for trees too deep for recursion. See Sum Root-to-Leaf Numbers.
System design
- Search over user statuses. Design text search over short posts that users write, with AND and OR queries and no ranking. The design was an inverted index: a map from each word to a sorted list of the status IDs that contain it. New statuses go through a queue that splits them into words and cleans them up (lowercase, word stems, dropping very common words) before the index is updated. AND means intersecting lists, OR means merging them, and skip pointers help jump ahead while intersecting. To grow past one machine, the discussion compared splitting the index by word, using consistent hashing, with splitting it by document and sending each query to every shard. Practise it as Design a Boolean Status Search Index.
Common questions
How many rounds is the Meta software engineer onsite?
Report 03 (level not given) described a phone screen with two coding tasks, then two onsite coding rounds of two tasks each and a system design round. The experienced report (01) and the new-grad report (02) list their onsite tasks without a round count.
Does the Meta onsite include system design?
Report 03 (level not given) had a system design round on AND/OR search over user statuses. The experienced report (01) and the new-grad report (02) named only coding tasks.
Do Meta onsite coding questions come with follow-ups?
For the experienced candidate (report 01), the stock task moved from unlimited trades to one trade. For the new grad (report 02), the palindrome task went from one deletion to k, the BST task added repeated range queries, and the vector task switched formats. For the candidate whose level wasn't given (report 03), the palindrome task also went to k deletions, and the tree task included a discussion of a loop-based walk.
Does Meta repeat onsite questions?
The palindrome-after-deletions task, one deletion first and then k, appears in both report 02 (new grad) and report 03 (level not given). No other task was repeated across these three reports.
Keep practicing
- Maximum Stock Profit with Transaction Limits: the unlimited-trades and one-trade versions from report 01.
- Mouse and Cheese Reachability: the probe-only search from report 01.
- Insert into a Sorted Circular List: the linked-list task from report 02.
- Minimum Time to Complete Target Courses: the course schedule with durations from report 02.
- Validate a Palindrome After Limited Deletions: the one task in both report 02 and report 03.
- Word Break with a Reusable Dictionary: the word-splitting task from report 02.
- Repeated BST Range Sum and Average: the three-step tree range task from report 02.
- Sparse Vector Dot Product from Tuples: the (index, value) pairs version from report 02.
- Find and Reconstruct a Binary-Matrix Path: the grid path from the report 03 phone screen.
- Count Subarrays with a Target Sum: the prefix-sum count from the report 03 phone screen.
- Find a Deterministic Local Minimum: the binary search from report 03.
- Query Whether an Island Size Exists: the island lookup from report 03.
- Sum Root-to-Leaf Numbers: the tree path sum from report 03.
- Design a Boolean Status Search Index: the search design round from report 03.
- Meta coding question bank: every Meta coding problem on FastPrep (sign-in needed).
- Meta system design bank: Meta design exercises (sign-in needed).
- Meta interview guide: FastPrep's Meta interview page.
- Meta hiring insights: FastPrep's Meta hiring insights page.
- Experience feed: more interview reports from other candidates.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.