← Back to Directory

Lyft interview notes, page 3 of 10

Large Enterprises

Most of these Lyft write-ups come from Software Engineers (41), then Product Managers (28) and Data Scientists (19). The most reported rounds are the technical phone screen (42), the product sense round (29), the system design round (12) and the behavioral / leadership round (4).

97 interview notes · updated Sep 2026

Lyft·Software Engineer·Technical Phone Screen

May 2026
Lyft software engineer interview with a file processing problem that had more edge cases than I expected. The question itself sounds simple but the follow-up constraints are where they actually test you.
  • Given a large UTF-8 text file, write a program that reads the K-th non-empty line without loading the entire file into memory. How do you handle edge cases like fewer than K non-empty lines, blank lines, and trailing newlines?

“The core read-and-count loop took me maybe five minutes to sketch out.”

View Post

Lyft·Software Engineer·Technical Phone Screen

May 2026
Lyft software engineer coding round, two questions back to back. Nothing too exotic but the second one had a twist I wasn't expecting where they wanted you to actually compare two designs out loud instead of just picking one and moving on.
  • Implement an LRU cache with a fixed capacity. It should support get and put operations, both running in O(1) average time, with least-recently-used eviction when the cache is full.
  • Build a simple in-memory key-value store supporting put, get, and delete. Then propose and compare a second implementation with different trade-offs, explaining the pros and cons of each approach.

“Classic problem, I've done this before, and yet I still fumbled the part where you have to keep the doubly linked list and the hashmap in sync on every operation.”

View Post

Lyft·Software Engineer·Onsite - Coding / Algorithms

May 2026
Lyft software engineer coding round, laptop-based, 1.5 hours to build an in-memory key-value store with transaction support. Time was tight and I did not get it fully bug-free.
  • Build an in-memory key-value database that supports begin, commit, and rollback for transactions.

“Ran out of time before I had a clean solution.”

View Post

Lyft·Product Manager·Onsite - Product Sense / Strategy

May 2026
Lyft PM interview with a product strategy question about launching related products. Pretty thin on details but the core question was meaty enough to chew on for a while.
  • As a PM, how would you approach launching a suite of related products?

“This is the kind of question where you can spiral fast if you don't anchor on something concrete first.”

View Post

Lyft·Data Scientist·Technical Phone Screen

May 2026
Lyft data scientist interview that leaned heavily on foundational stats, the kind of stuff you think you know cold until someone asks you to explain it to a room full of non-technical stakeholders.
  • State Bayes' theorem and walk through a concrete example showing how it works.
  • Explain what a p-value actually means in plain terms, including what it does NOT tell you.

“I knew the formula but fumbled the intuition part.”

View Post

Lyft·Software Engineer·Technical Phone Screen

May 2026
Lyft software engineer round that was basically a deep dive into a single system design problem. More involved than I expected for what felt like a coding screen at first.
  • Design a key-value store that supports set(key, value, timestamp) and get(key, timestamp), where get returns the value at the largest stored timestamp that doesn't exceed the query timestamp. Handle up to 100k keys and 1 million operations, and discuss data structure choices, memory growth, and persistence via serialization.

“I started with a sorted map per key and binary search, which is the right instinct, but fumbled a bit explaining why sorted insertion is still O(log n) amortized versus a naive array.”

View Post

Lyft·AI Engineer·Onsite - System Design / Architecture

May 2026
System design round at Lyft for an AI Engineer role. One big open-ended question about monoliths vs microservices that sprawled into like six different sub-topics. Felt like a PhD dissertation prompt squeezed into 45 minutes.
  • Walk through the design trade-offs between building as a single monolithic service versus decomposing into microservices. Cover when each makes sense, how you'd scale and deploy them, developer experience, how to migrate if needed, data ownership, and observability costs.

“This question has so many branches that I spent the first few minutes just trying to figure out where to start.”

View Post

Lyft·Software Engineer·Technical Phone Screen

May 2026
Lyft data science interview with a probability question that sounds deceptively simple but trips you up if you rush the setup.
  • N riders each receive a discount coupon, and each rider independently uses their coupon with probability P. What is the probability that at least one coupon gets used?

“I jumped straight to 1 minus (1-P)^N which is right, but I fumbled explaining why for a second.”

View Post

Lyft·Product Manager·Onsite - Product Sense / Strategy

May 2026
Lyft product design interview, one question about building an ad product. Pretty open-ended and I wasn't totally sure how deep to go on the monetization side versus the user experience side.
  • Design an advertising product for Lyft.

“I started with the rider experience because I figured ads in a captive in-car context are different from, say, a feed.”

View Post

Lyft·Machine Learning Engineer·Technical Phone Screen

May 2026
Lyft ML Engineer interview that was basically a deep dive into one coding problem. They wanted a full one-hot encoder from scratch, no libraries, and then kept pulling the thread on edge cases and scalability until I ran out of things to say.
  • Build a one-hot encoder class from scratch (no sklearn or pandas) with fit, transform, and fit_transform methods.
  • How should the encoder handle categories at transform time that were never seen during fit? Walk through the tradeoffs of each approach.
  • What's the time and space complexity of your encoder, and how would you scale it to columns with very high cardinality?
  • Implement an inverse_transform method that maps encoded vectors back to their original category labels.

“I started with fit and just stored sorted unique categories in a list.”

View Post