I interviewed for Samsung R&D (SRI-D) and cleared all three rounds. Here's what each looked like. ## Round 1: Aptitude Mixed bag of questions covering: - Train crossing problems - Blood relations and direction-based puzzles - Work done by X and Y individually, then together - Number sequence prediction Nothing too tricky, standard aptitude prep covers it. ## Round 2: Technical Round 1 **OOPs** — polymorphism (static vs dynamic, overloading vs overriding), virtual functions > A virtual function enables runtime polymorphism — the correct function is resolved at runtime via vtable lookup, not at compile time. **DSA** Number of Islands: ```cpp #include using namespace std; void dfs(vector>& grid, int i, int j) { if (i < 0 || i >= grid.size() || j < 0 || j >= grid[0].size() || grid[i][j] != '1') return; grid[i][j] = '0'; dfs(grid, i+1, j); dfs(grid, i-1, j); dfs(grid, i, j+1); dfs(grid, i, j-1); } int numIslands(vector>& grid) { int count = 0; for (int i = 0; i < grid.size(); i++) for (int j = 0; j < grid[0].size(); j++) if (grid[i][j] == '1') { dfs(grid, i, j); count++; } return count; } ``` Merge two arrays without duplicates: ```cpp #include #include using namespace std; vector mergeWithoutDuplicates(vector& a, vector& b) { vector merged; for (int x : a) merged.push_back(x); for (int x : b) merged.push_back(x); sort(merged.begin(), merged.end()); vector result; for (int i = 0; i < merged.size(); i++) if (i == 0 || merged[i] != merged[i-1]) result.push_back(merged[i]); return result; } ``` Can this be done in-place? Technically yes if both arrays are already sorted — use two pointers and write back. But without sorting, you need O(n) space to track seen elements. Worth clarifying in the interview. Sorting — worst case complexities and best use cases: - **Bubble** — O(n²) — nearly sorted small arrays - **Insertion** — O(n²) — small or nearly sorted data - **Selection** — O(n²) — when writes are expensive (minimizes swaps) - **Quick** — O(n²) worst, O(n log n) avg — general purpose large arrays - **Heap** — O(n log n) — when guaranteed O(n log n) is needed **ML** — identifying supervised learning from a list **Networking** — OSI layers **OS** — system calls and what they trigger ## Round 3: Technical Round 2 **REST API** — HTTP methods (GET, POST, PUT, PATCH, DELETE) and their differences, server error codes (5xx vs 4xx, common ones like 500, 502, 503, 404, 401, 403) **CNN basics** — architecture, convolution operation, pooling, how filters learn features **Portfolio question** — difference between dB and LUFS > dB is a general unit for relative signal level. LUFS (Loudness Units Full Scale) is a perceptual loudness measure that accounts for how humans hear — it integrates loudness over time using ITU-BS.1770 weighting. dB SPL tells you amplitude, LUFS tells you perceived loudness.

My Blog