OpenAI Interview Questions — Page 38

OpenAI·Software Engineer·Technical Phone Screen

May 2026

Live coding round at OpenAI for a SWE role, basically build a ChatGPT-style chat UI in React from scratch during the interview. Pretty intense scope for a single session, covering streaming, state management, error handling, and cleanup all at once.

Implement a ChatGPT-like chat UI in React that supports a message list with user and assistant roles, a text input with enter-to-send behavior, a loading state while waiting for the assistant, and streaming token-by-token responses rendered incrementally.
How would you handle rapid consecutive sends or multiple button clicks while a response is still streaming?
How do you clean up an in-flight streaming request when the component unmounts?
How would you handle a failed API response in this chat UI, including giving the user a way to retry?

“This is a lot to hold in your head at once.”

View Post

OpenAI·Software Engineer·Onsite - System Design / Architecture

May 2026

Went through a system design round at OpenAI for a software engineer role. The focus was on distributed system reliability, specifically around failure handling in a multi-node setup. Pretty deep dive into the details.

In a distributed system where nodes can crash mid-operation, how do you prevent retries from causing duplicate processing or double-counting?

“This was a follow-up and it's where things got interesting.”

View Post

OpenAI·Software Engineer·Onsite - System Design / Architecture

May 2026

System design round at OpenAI for a software engineering role. The whole thing was basically one long deep-dive into a key-value store, starting from a simple single-file design and getting progressively messier with multi-file storage, concurrency, and crash recovery. Pretty intense but not unfair.

You have a basic persistent key-value store with put, get, and delete operations backed by a single append-only file and an in-memory index. Extend it to support multiple size-capped data files with rollover.
How would you handle compaction across multiple files in this design?
Make the store safe for concurrent reads and writes. What locking strategy would you use?
How do you make the in-memory index durable across restarts? What are the tradeoffs between rebuilding from scratch versus persisting a snapshot?
How should the compactor run without blocking readers, and how do you safely swap in the new compacted file?
How do you detect and recover from a partial write at the tail of a file after a crash?

“This is where most of the time went.”

View Post

OpenAI·Software Engineer·Technical Phone Screen

May 2026

Interviewed for a SWE role at OpenAI and got a pretty deep ML systems question right out of the gate. The whole thing was focused on implementing multi-head attention from scratch, shapes and all. Not a vibe check, they wanted real code and real reasoning.

Implement the forward pass of Multi-Head Attention using NumPy or PyTorch. You need to handle splitting into heads, scaled dot-product attention with an optional mask, concatenating outputs, and explaining how you keep the softmax numerically stable. Clearly state tensor shapes throughout.

“This was the whole interview basically.”

View Post

OpenAI·Software Engineer·Onsite - System Design / Architecture

May 2026

System design round at OpenAI for a software engineer role. The question was one massive prompt covering basically every dimension of a GPT-facing web app you can think of, streaming, snapshots, search, sharing, auth, scaling, testing, the whole thing. Left feeling like I'd only scratched the surface on half of it.

Design an end-to-end web application for interacting with a GPT-like model that supports real-time token streaming, snapshot saving, full-text search over saved sessions, and link-based sharing with role-based access controls. Walk through the full architecture including frontend, backend, data stores, APIs, and all the operational concerns.

“This was the entire interview, one question, no breaks.”

View Post

OpenAI·Software Engineer·Onsite - System Design / Architecture

May 2026

System design round at OpenAI for a software engineer role. The whole session was basically one big chess platform question that kept branching into more and more sub-problems. Dense but interesting.

Design a real-time online chess platform. Walk through requirements, core entities, real-time move sync, game state persistence, concurrency, spectator support, matchmaking, and how you'd scale the system.

“This question just kept expanding.”

View Post

OpenAI·Software Engineer·Technical Phone Screen

May 2026

Interviewed for a software engineer role at OpenAI and got a grid simulation problem that was sneakily harder than it looked on the surface. The BFS angle was obvious but tracking the recovery/immunity timing tripped me up for a bit.

Given a grid where each person starts healthy, infected, or immune: infected people spread to adjacent healthy neighbors each day, and after D days an infected person recovers and becomes permanently immune. How many days until no more state changes are possible?

“My first instinct was plain BFS and I started coding before fully thinking through the immunity part.”

View Post

OpenAI·Machine Learning Engineer·Technical Phone Screen

May 2026

Interviewed for an ML Engineer role at OpenAI. Two very dense technical questions, both requiring serious depth. Left feeling like I'd done okay on the first one but probably exposed some gaps on the from-scratch implementation part.

You're given a CIFAR-scale image dataset with label noise, corrupted samples, and class imbalance. How do you build a baseline classifier and a full data quality improvement plan, including detecting noise, filtering bad samples, handling imbalance, splitting data safely, and comparing mitigation strategies like confidence-based pruning, co-teaching, and augmentations like MixUp or CutMix?
Using only NumPy-style linear algebra, implement forward and backward passes for a two-layer network with ReLU and softmax cross-entropy from scratch. Derive the gradients analytically, validate with numerical gradient checks, and discuss numerical stability, weight initialization, regularization, and how you'd extend this to a CNN architecture.

“This one I actually felt decent about.”

View Post

OpenAI·Machine Learning Engineer·Technical Phone Screen

May 2026

Interviewed for an ML Engineer role at OpenAI and got a binary search problem dressed up as a dependency versioning question. Felt straightforward at first but the edge case discussion is where things got interesting.

Given a sorted list of software versions and a predicate that checks if a version satisfies a dependency, find the first (and/or last) working version. Working versions are contiguous in the sorted order. What's the optimal approach and how do you verify its correctness?

“I recognized it as binary search pretty fast, which felt good.”

View Post

OpenAI·Software Engineer·Onsite - System Design / Architecture

May 2026

System design round at OpenAI for a software engineer role. The whole session was basically one massive question about building a cloud IDE from scratch, and they kept pushing deeper on every subsystem I touched.

Design a cloud-based IDE (similar to Codespaces or Replit), covering containerized workspace provisioning per user, file system persistence, low-latency collaborative editing, terminal and language server multiplexing, build/run sandboxing, and cost management for idle workspaces.

“This was a lot.”

View Post