Palantir Software Engineer Onsite Coding Questions

During the Palantir virtual onsite, I failed to complete the technical assessment within the time limit. I pursued an overly complex approach instead of a standard recursive solution, which prevented

This was the rejection email I got after the final round palantir internship interviews. I thought that all of the technical, decomp and learnings interviews went really well and I got on really well

Has anyone gone through the virutal-onsite at Palantir before? I would like to know what kind of questions were asked in the learning segment of the interview. How does one even prepare for this? The

Hi everyone, I have a the final HM round for palantir for their new grad fdse role. I was wondering if anyone has gone through that process and has any tips for it. I heard they do reject a lot after

LeetCode #3006: Find Beautiful Indices in the Given Array I. Difficulty: Medium. Topics: Two Pointers, String, Binary Search, Rolling Hash, String Matching, Hash Function. Asked at Palantir in the last 6 months.

LeetCode #443: String Compression. Difficulty: Medium. Topics: Two Pointers, String. Asked at Palantir in the last 6 months.

LeetCode #244: Shortest Word Distance II. Difficulty: Medium. Topics: Array, Hash Table, Two Pointers, String, Design. Asked at Palantir in the last 6 months.

LeetCode #1202: Smallest String With Swaps. Difficulty: Medium. Topics: Array, Hash Table, String, Depth-First Search, Breadth-First Search, Union-Find, Sorting. Asked at Palantir in the last 6 months.

I recently had my coding interview with them and now got moved to the virtual onsite. It is 2 back to back interviews. Does anyone have any tips for the decomposition / live learning rounds? If anyone

Hi Everyone, I have the palantir virtual onsite coming up and it's two interview which are their decomp and learning interviews. I was wondering if any of you guys had any insight on what to expect fr

Just wondering how many are waiting for either an interview (onsite, HM interview) or an offer for Palantir’s Path program

## Problem You have a tree of resources (like a file system). Each node can have explicit permissions set for users. If a node has no explicit permission for a user, it inherits from its parent. An explicit `DENY` always overrides an inherited `ALLOW`. Implement: ```python class AccessControlTree: def __init__(self): ... def add_node(self, node_id: str, parent_id: str | None) -> None: ... def set_permission(self, node_id: str, user: str, perm: str) -> None: # perm: "ALLOW"|"DENY" def can_access(self, node_id: str, user: str) -> bool: ... ``` ## Example ``` tree.add_node("root", None) tree.add_node("docs", "root") tree.add_node("secret", "docs") tree.set_permission("root", "alice", "ALLOW") tree.set_permission("secret", "alice", "DENY") tree.can_access("docs", "alice") -> True # inherits from root tree.can_access("secret", "alice") -> False # explicit DENY overrides tree.can_access("root", "bob") -> False # no permission set anywhere ``` ## Follow-ups 1. How does your traversal strategy change if the tree is very deep (10,000+ levels)? 2. What if a user can belong to multiple groups, each with their own permissions? 3. How would you batch-evaluate access for 1,000 users across 500 nodes efficiently?

## Problem Group a list of words into anagram clusters, returning a dictionary mapping sorted keys to word lists. ## Likely LeetCode equivalent Directly matches Group Anagrams (LC 49). ## Tags hash_table, strings, sorting, palantir

## Problem Design a spatial index that supports storing 2D points and efficiently answering proximity queries. Implement: ```python class SpatialIndex: def insert(self, point_id: str, x: float, y: float) -> None: ... def delete(self, point_id: str) -> None: ... def nearest(self, x: float, y: float, k: int) -> list[str]: ... def range_query(self, x1: float, y1: float, x2: float, y2: float) -> list[str]: ... ``` `nearest` returns the `k` closest point IDs by Euclidean distance. `range_query` returns all points within the given bounding box. ## Example ``` index.insert("A", 1.0, 2.0) index.insert("B", 3.0, 4.0) index.insert("C", 1.5, 2.5) index.nearest(1.0, 2.0, k=2) -> ["A", "C"] # closest two to (1,2) index.range_query(0, 0, 2, 3) -> ["A", "C"] # within bounding box ``` ## Follow-ups 1. Compare a grid-bucket approach vs. a k-d tree for this workload. When does each win? 2. How do you handle points clustered in a small area causing one bucket to dominate? 3. What changes if points are moving (e.g., tracking vehicles in real time)? 4. How would you scale this index to billions of points across multiple machines?