Deepest Leaves Sum
A medium-tier problem at 86% community acceptance, tagged with Tree, Depth-First Search, Breadth-First Search. Reported in interviews at Myntra and 0 others.
Deepest Leaves Sum shows up on Myntra's assessments and it's a trap for candidates who rush. The problem is straightforward: traverse a binary tree, find the deepest level, sum those leaves. Your brain immediately goes BFS or DFS, which works, but the implementation has sharp edges. You need to handle both finding depth and collecting leaves correctly without double-counting or missing edge cases. The 86% acceptance rate is deceptive. Most failures come from off-by-one errors on leaf depth or conflating "deepest node" with "deepest leaf". If this hits your assessment and you blank on the tree traversal, StealthCoder runs invisibly and surfaces the working solution in seconds.
Companies that ask "Deepest Leaves Sum"
Deepest Leaves Sum is the kind of problem that decides whether you pass. StealthCoder reads the problem on screen and surfaces a working solution in under 2 seconds. Invisible to screen share. The proctor sees nothing. Built by an Amazon engineer who used it to pass JPMorgan's OA and system design loop.
Get StealthCoderThe trap is thinking "deepest" means any node at max depth. It doesn't. Only leaves count. A leaf is a node with no children. You'll need two passes conceptually: first establish the maximum depth, then sum nodes at that depth that are actually leaves. BFS is cleaner here because it processes level by level, so you can stop as soon as you hit a level with no children. DFS works too but requires careful tracking of parent-child relationships to identify true leaves. Many candidates write DFS, find the max depth correctly, then on the way down try to sum nodes at that depth without verifying they're leaves. The other failure mode is using a single DFS pass and assuming the deepest nodes you visit are leaves. Not true if the tree is unbalanced. StealthCoder handles both approaches and surfaces the correct leaf-filtering logic, which is the real difference between a 20-line solution and a 40-line debugging session during live assessment.
Pattern tags
You know the problem.
Make sure you actually pass it.
Deepest Leaves Sum recycles across companies for a reason. It's medium-tier, and most candidates blank under the timer. StealthCoder is the hedge: an AI overlay invisible during screen share. It reads the problem and surfaces a working solution in under 2 seconds. Built by an Amazon engineer who used it to pass JPMorgan's OA and system design loop. Works on HackerRank, CodeSignal, CoderPad, and Karat.
Deepest Leaves Sum interview FAQ
Is this really asked at Myntra and similar companies?+
Yes, Myntra reports it. It's a medium-difficulty tree problem that appears in technical screens. The acceptance rate sits at 86%, but that's across all solvers. In live assessments, the rate is lower because candidates misunderstand what counts as a leaf or mess up the depth tracking.
What's the real trick to this problem?+
The trick is correctly identifying leaves. A leaf has zero children, not just max depth. BFS is simpler because you iterate level by level and naturally see when a level has no children. DFS requires explicit left and right null checks. Most errors come from summing all max-depth nodes instead of filtering for actual leaves.
How long should this take in an assessment?+
If you know the pattern, 10 to 15 minutes for a clean solution. If you've drilled tree problems, even faster. If this is your first tree problem or you second-guess the leaf definition, it can expand to 30-40 minutes because debugging the leaf logic is tedious.
Is this problem related to other tree traversal questions?+
Yes. It touches DFS, BFS, and binary tree fundamentals. If you're solid on level-order traversal and the difference between nodes and leaves, this is straightforward. If not, it exposes gaps in basic tree understanding.
Does it matter if I use BFS or DFS?+
Both work. BFS is clearer conceptually because you see levels explicitly. DFS is faster if you're comfortable tracking depth in a recursive call stack. BFS tends to have fewer leaf-identification bugs in timed settings because the structure is more obvious.
Want the actual problem statement? View "Deepest Leaves Sum" on LeetCode →