← Back to Directory

Harvey interview notes, page 2 of 6

Mid-sized

Most of these Harvey write-ups come from Software Engineers (46), then Backend Engineers (4) and Machine Learning Engineers (4). The most reported rounds are the technical phone screen (30), the system design round (17), the behavioral / leadership round (4) and the onsite coding round (2).

54 interview notes · updated Jul 2026

Harvey·Software Engineer·Onsite - System Design / Architecture

Apr 2026
System design round at HarveyAI for a software engineer role. The prompt was a data room product, basically Google Drive but PDF-only, with the main focus being ACL design at the organization level. Niche enough that I wasn't sure how deep to go on the storage layer versus the permissions model.
  • Design a data room product similar to Google Drive but restricted to PDFs by default. Focus specifically on how you'd implement access control lists at the organization level.

“I spent too long on the file storage side and had to scramble when they steered me toward ACLs.”

View Post

Harvey·Backend Engineer·Onsite - System Design / Architecture

Apr 2026
System design round at Harvey for a backend role. The whole thing centered on building a secure document vault, which sounds straightforward until you get into the weeds on auth, chunked uploads, and integrity verification. Pretty thorough for a single question.
  • Design a secure cloud-based document vault where authenticated users can upload, store, and retrieve sensitive files with strict authorization controls.
  • How would you scale this system as the number of files and user traffic grows significantly?
  • How would you design the access control and permission model for this system?
  • How would you handle uploading very large files in this system?
  • When files are uploaded to object storage in chunks using presigned URLs, how do you verify that each chunk actually belongs to the intended file and was uploaded by the authorized user, rather than being arbitrary or malicious data?

“This was the anchor question for the whole session.”

View Post

Harvey·Machine Learning Engineer·Technical Phone Screen

Apr 2026
Technical round at Harvey AI for an ML engineer role, centered on a pretty involved NLP-adjacent problem. No ML libraries allowed which made it more interesting than I expected.
  • Given an input string, design a tree or graph-based algorithm to transform it into a target tokenized sequence. You must simulate embeddings without using any ML models, and validate the final output.

“This one took me a minute to even parse what they were asking.”

View Post

Harvey·Software Engineer·Technical Phone Screen

Apr 2026
Harvey AI coding round, one big design question that took the whole session. The problem looked like a toy filesystem at first but the constraints turned it into a mess of edge cases I was not fully prepared for.
  • Design and implement an in-memory hierarchical file system with addFile and get operations, a per-directory capacity limit of 5 entries, and OS-style auto-renaming for duplicate file names within the same directory.

“The rename logic is where I lost time.”

View Post

Harvey·Software Engineer·Technical Phone Screen

May 2026
Interviewed at HarveyAI for a software engineering role and got a pretty interesting coding problem about text highlighting with multiple sources. Two parts to it, base case first then a follow-up that added priority ordering. Not a typical leetcode grind question which was kind of refreshing.
  • Given a text string and multiple sources each providing non-overlapping highlight intervals, implement a function that applies all highlights to the text.
  • Now extend the solution: sort sources by how many highlight segments they have (most segments first), and when applying highlights in that order, skip any portion of a new segment that overlaps with already-highlighted text.

“The base case wasn't too bad.”

View Post

Harvey·Software Engineer·Onsite - System Design / Architecture

May 2026
Got a system design coding question at Harvey for a software engineer role. The problem was building an in-memory spreadsheet engine from scratch, which sounds manageable until you get into propagation and cycle detection.
  • Design and implement an in-memory spreadsheet engine that supports named cells, integer literals, and formulas that sum cell references. get_cell must be O(1) using cached values, and set_cell must propagate updates to all dependent cells.
  • How do you detect and handle dependency cycles in the spreadsheet engine, and how do you ensure the engine stays in a valid state if a set_cell call is rejected?

“The caching requirement is what trips you up first.”

View Post

Harvey·Software Engineer·Technical Phone Screen

May 2026
Coding screen for a software engineer role at HarveyAI. One question, pretty focused on string parsing logic. Nothing crazy but the whole-word matching constraint is where people trip up.
  • Given a list of sentences and a list of tag words, count how many sentences contain at least one tag as a whole word (case-insensitive, no substring matches).

“The substring trap is real here.”

View Post

Harvey·Software Engineer·Onsite - System Design / Architecture

May 2026
System design round at HarveyAI for a software engineer role, focused entirely on building a RAG pipeline from ingestion to answer generation. Dense technical territory and they clearly wanted you to go deep on every layer, not just wave at the problem.
  • Design a retrieval-augmented generation system that grounds an LLM's responses in an external knowledge base, covering ingestion, retrieval, and answer generation end to end.
  • How would you handle document ingestion from multiple source types like PDFs, web pages, and internal wikis, and what chunking strategy would you use?
  • Walk me through how you'd combine dense vector search with sparse retrieval like BM25, and how reranking fits in.
  • How would you evaluate the quality of a RAG system, both at the retrieval layer and the final answer quality?
  • What operational concerns would you flag for a RAG system handling sensitive enterprise data, particularly around access control and PII?

“This is a beast of a question.”

View Post

Harvey·Software Engineer·Onsite - Coding / Algorithms

May 2026
Coding round at HarveyAI for a software engineer role, one algorithmic problem centered on text highlighting with overlapping ranges. Pretty focused interview, no fluff.
  • Given a string and a list of highlight ranges that may overlap, produce the final highlighted output by merging overlapping ranges into contiguous spans. How would you approach this, and what's the time complexity?

“The core idea is a sweep-line interval merge, sort the ranges and walk through them collapsing overlaps.”

View Post

Harvey·Software Engineer·Technical Phone Screen

May 2026
Got a coding question at Harvey for a software engineer role that was basically a string manipulation problem with a merging twist. Not the hardest thing I've seen but the follow-ups added some real complexity.
  • Given a string and a list of source strings, find every exact (case-sensitive) occurrence of any source string in the main text, wrap matched characters with highlight tags, and merge any overlapping or adjacent matched ranges into a single highlighted region before returning the final string.
  • As a follow-up: for each source string, count how many exact occurrences it has in the original text.
  • Second follow-up: for each merged highlighted region in the output, return which source strings contributed at least one match inside that region.

“The merging part is what gets you.”

View Post