SYSTEM DESIGN

NVIDIA System Design Questions

7 distinct system design problems asked at NVIDIA from September 2025 to September 2026, merged from 7 candidate reports. Each one links to a full solution.

Sources: PracHub, LeetCode discussions. Sorted by how often the problem was reported at NVIDIA.

Design an Autonomous-Vehicle Perception Pipeline

Design the camera perception pipeline for an autonomous vehicle: capture from multiple cameras, time synchronization, calibration, preprocessing, neural network inference (object detection, segmentation), sensor fusion and…

HLD1 report·last Aug 2026

Design the Control Plane for a Compute Cluster: Host Monitoring + Job Dispatch

Design the control plane for a compute fleet of about 100,000 GPU hosts. Each host continuously reports health (GPU ECC errors, temperature, NVLink status, heartbeat); the control plane keeps live host state under high write…

HLD1 report·last Jun 2026·also asked at 1 other company

Design a GPU Telemetry Collection and Analytics Platform

Design a telemetry pipeline for a very large fleet of devices or sensors (e.g., millions of temperature sensors, GPUs, or IoT devices each reporting every few seconds). Devices may be offline and upload batched, late-arriving…

HLD1 report·last May 2026·also asked at 4 other companies

Design a URL shortening service

Design a URL shortening service like TinyURL or bit.ly. Users submit a long URL and get a short code (optionally a custom alias and expiry); visiting the short URL redirects to the original. Assume hundreds of millions of new…

HLD1 report·last Feb 2026·also asked at 7 other companies

Design a bidirectional data sync dashboard

Design a dashboard system that keeps data synchronized in both directions between the dashboard's own store and one or more external systems: changes made in the dashboard propagate to the external systems and vice versa. Discuss…

HLD1 report·last Feb 2026

HashMap without STL

Implement a hash map from scratch without standard library containers (e.g., in C/C++): put, get and remove with average O(1) time. Discuss the hash function, collision handling (separate chaining vs open addressing), load factor…

LLD1 report·last Feb 2026

Design first-time Kubernetes deployment in new cloud

Design how you would do a first-time Kubernetes deployment of your services in a new cloud environment: account/project bootstrapping, networking (VPC, subnets, ingress, load balancers), cluster setup and node pools (including…

HLD1 report·last Jan 2026

Practice with a Mock Interview

Apply these questions in a live system design mock interview.

Start System Design Interview →