What is Interview Forge?

Interview Forge helps you practice mock interviews, follow study plans, track career goals, review learning analytics, and prepare across 36 pages.

Design a Typeahead Suggestion Service at NVIDIA — System Design Interview Guide

The problem at NVIDIA

Functional: as the user types, return the top-k suggestions matching the prefix.
Non-functional: <50ms response, personalization optional, updates as trends change.
Scale: ~1B queries/day, ~100M distinct prefixes.
Consistency: suggestions can be slightly stale (trending updates in seconds); personalization is best-effort.

Out of scope (covered in detail below)

Spell correction, full search ranking, multi-language/emoji, and trending detection — deferred and discussed in the dedicated section.

At NVIDIA, this problem is framed around systems, performance, and parallel thinking. Expect NVIDIA-specific scale numbers and a focus on production readiness and the company's engineering bar.

Capacity estimation

Resource	Estimate	Notes
Queries/day	~1B	Keystroke-driven, very chatty
Distinct prefixes	~100M	Indexable
Response time	<50 ms	Per keystroke
Top-k	~10	Per prefix

Architecture overview

Index + cache + scorer

+--------+ keystroke +----------------+  prefix  +-----------+
| Client |---------->| Suggestion Svc |--------->| Index     |
+--------+           +----------------+          | (trie/n-gram)|
                           |                    +-----------+
                           v                          ^
                     +-----------+   top-k            |
                     | Redis     |<-------------------+
                     | (per prefix)
                     +-----------+
                           ^
                           |
                     +-----------+  query logs
                     | Batch     |<--------- (offline)
                     | Scorer    |
                     +-----------+

A prefix index maps to candidates; a batch job rebuilds scores; a cache absorbs the read load.

Step-by-step request flow

On each keystroke, the client sends the current prefix to the suggestion service (debounced on the client).
The service checks a per-prefix cache (Redis) for the top-k; on a hit, return immediately.
On a miss, query the prefix index (trie or n-gram) for candidates, sort by score, take top-k, and cache.
A batch job rebuilds scores from query logs; a real-time path handles trending terms.

Unlock the full guide

Premium members get every DSA, system design, company and role interview guide — plus AI practice, mock interviews, and analytics.

Unlock with Premium