"Practice LeetCode" is useless advice once you have an interview date. Amazon, Google, and Microsoft sample the same data structures with different weights, and a fresher with three weeks left should be solving the patterns their specific company over-indexes on — not the same generic 150 problems everyone reposts. Below is the honest, company-by-company breakdown, mapped to topics from the curated 370 so you know exactly what to filter down to.

None of this is leaked questions — companies rotate those. It's the pattern distribution each is known for, which is far more stable and far more useful for placement prep.

Amazon — grids, graphs, trees, and the OA debug trap

Amazon is the most pattern-predictable of the three for freshers, which is good news. The Online Assessment (OA) and the interview loop lean hard on matrix/grid traversal, graphs, and tree problems, often dressed up as warehouse, delivery, or org-chart stories. Expect BFS/DFS on a grid, connected components, and tree recursion.

• Go deep: Graphs (BFS/DFS, number-of-islands style grid flooding, topological sort), trees (traversals, lowest common ancestor), and the occasional grid-based DP (minimum path, unique paths).
• The OA twist: Amazon's OA often includes a debugging / code-completion section — you fix broken code instead of writing from scratch. Practise reading unfamiliar code under time pressure, not just writing it.
• Don't ignore: Leadership Principles. Amazon weights behavioural answers heavily even for SDE-1, including in campus and off-campus loops. Have your STAR stories ready.

Google — intervals, greedy, and clean problem modelling

Google interviews are less about exotic data structures and more about how cleanly you model a fuzzy problem. The recurring technical lean is intervals, greedy, and arrays/strings with a twist — plus a strong expectation that you discuss complexity and edge cases out loud before you write a line of code.

• Go deep: Intervals (merge, insert, meeting rooms), greedy (activity selection, jump game), two pointers, and binary search — including the "search on the answer" variant Google leans on.
• The differentiator: Google scores communication as much as correctness. A working brute force you explained well often beats a silent optimal solution.
• Watch for: Follow-ups that change the constraints mid-problem. They want to see you adapt, not recite a memorised template.

Microsoft — arrays, strings, and solid fundamentals

Microsoft is the most fundamentals-first of the three. For freshers, the bulk of questions sit in arrays, strings, linked lists, and recursion, with trees and DP showing up at the harder end. It rewards clean, bug-free code over cleverness.

• Go deep: Arrays & strings (in-place manipulation, sliding window), linked lists (reversal, cycle detection, merging), recursion, and basic trees.
• Reach topics: 1D/2D dynamic programming and backtracking appear, but rarely the segment-tree-tier stuff.
• What they grade: Edge-case handling and readable code. Microsoft interviewers notice missing null checks and off-by-one errors.

The pattern map at a glance

Company	Weights heavily	Lighter	The catch
Amazon	Grids, graphs, trees, grid-DP	Advanced math	OA debug section + Leadership Principles
Google	Intervals, greedy, binary search	Heavy data-structure trivia	Communication is scored
Microsoft	Arrays, strings, linked lists, recursion	Exotic structures	Clean, edge-case-safe code

How to actually use this

Don't grind all 370 problems hoping the right ones come up. Pick your target company, and drill the patterns it over-weights until you recognise them in under two minutes, then breadth-fill the remaining topics. An Amazon-bound fresher should have grid and graph problems on autopilot before touching, say, bit manipulation. A Microsoft candidate should have arrays and linked lists bulletproof first.

This is the exact reasoning SUITS automates: it reorders the curated sheet so your highest-frequency patterns float to the top instead of being buried at step 14. For a role-level version of this — frontend vs backend vs SDE — read the DSA roadmap by role. And if you want the universal frequency ranking that underlies all three companies, the 21 topics that cover 90% of interviews are the foundation to anchor on before you specialise.