Discord Interview Guide 2026: Real-Time Messaging, Elixir + Rust, Voice Infrastructure, and Trust & Safety

Discord Interview Process: Complete 2026 Guide

Overview

Discord is the communications platform originally built for gamers that has grown into a general-purpose community and communication tool with ~200M monthly active users across millions of servers in 2026. Founded 2015 by Jason Citron and Stanislav Vishnevskiy, private with a 2021 valuation of $15B and subsequent tender offers. ~1,200 employees in 2026 after 2023–2024 rightsizing cycles that narrowed focus back to core communication and community product. Headquartered in San Francisco with significant remote engineering presence. The platform handles real-time messaging, voice / video, community management, activities (in-channel games and applications), and increasingly creator and commerce features. Discord is famous in engineering circles for its Rust-backend rewrite of performance-critical services (from Go / Elixir originally), its aggressive use of Elixir for real-time messaging at scale, and publicly-written-about engineering work on message storage (moving from Cassandra to ScyllaDB) and real-time infrastructure. Interviews reflect the distinctive combination of real-time-systems depth, consumer-scale reality, and gaming-adjacent culture.

Interview Structure

Recruiter screen (30 min): background, why Discord, team preference. The engineering surface: messaging infrastructure (Elixir / Rust), voice / video (WebRTC, custom stack), client (React / TypeScript on web, Swift / Kotlin on mobile, Electron for desktop), data / ML (safety, moderation, recommendations), platform (activities, commerce), and infrastructure.

Technical phone screen (60 min): one coding problem, medium-hard. Python for infrastructure / tools; TypeScript for web client; Rust for performance-critical services; Elixir for messaging; Swift / Kotlin for mobile. Problems tilt applied — implement an event handler, build a pub/sub primitive, handle a streaming update.

Take-home (some senior / staff roles): 4–6 hours on a realistic engineering problem.

Onsite / virtual onsite (4–5 rounds):

• Coding (1–2 rounds): one algorithms round, one applied systems round. The applied round often involves real-time primitives — implement a chat-message fan-out, build a connection pool for WebSocket clients, handle a streaming event with backpressure.
• System design (1 round): real-time messaging / consumer-scale prompts. “Design the message-fanout system delivering 1M messages/sec to users across millions of channels.” “Design the voice-routing system for 10K simultaneous multi-user voice calls per region.” “Design moderation workflows for user-reported content at 200M MAU scale.”
• Domain / real-time-systems deep-dive (1 round): WebSocket / gateway architecture, pub/sub at scale, eventual-consistency semantics for message history, voice-media routing, or Elixir / Rust idioms depending on role.
• Product / user round (1 round): distinctive at Discord. Discussion of Discord as a product, community dynamics, creator / server-owner UX. Candidates who use Discord deeply have a clear edge.
• Behavioral / hiring manager: past projects, real-time systems experience, consumer-product empathy.

Technical Focus Areas

Coding: depending on target team, different language focus. Python for tools / data; TypeScript for web; Rust for high-performance services (Discord has actively rewritten performance bottlenecks in Rust); Elixir for messaging backend; Swift / Kotlin for mobile.

Real-time messaging: WebSocket gateway at massive scale, connection management for millions of concurrent clients, pub/sub fan-out, delivery guarantees under at-least-once semantics, client-reconnection handling, presence and awareness.

Elixir / OTP: Discord runs substantial Elixir for real-time messaging. OTP fundamentals (GenServer, Supervisor, Registry), actor-model intuition, distributed-Erlang clustering, monitoring and observability for BEAM processes.

Rust: high-performance services in Rust where Elixir / Go don’t meet latency / throughput needs. Discord’s famous “Why Discord is switching from Go to Rust” blog post documents their read-states service rewrite; similar rewrites continued for other bottlenecks.

Storage: Discord’s move from Cassandra to ScyllaDB for message storage is widely discussed. For data / infrastructure roles, understanding this and the reasoning (tail latency, operational overhead) matters.

Voice / video: custom SFU for voice, WebRTC for client connections, codec integration (Opus for voice), NAT traversal, regional routing for latency optimization.

Client engineering: React / TypeScript for web and desktop (Electron), native iOS / Android, handling thousands of real-time events per session without UI degradation.

Safety / moderation: content moderation at 200M MAU scale, reporting workflows, ML classifiers for CSAM detection and abuse patterns, appeals processes, Trust & Safety engineering is a significant investment area.

Coding Interview Details

Two coding rounds, 60 minutes each. Difficulty is medium-hard. Comparable to Cloudflare or HashiCorp — below Google L5 on pure algorithms, higher on real-time / concurrent systems.

Typical problem shapes:

• Implement a publish / subscribe primitive with fan-out semantics
• Build a connection manager handling many concurrent WebSocket-like connections
• Parse a streaming event log with backpressure and ordering guarantees
• Implement a rate limiter or circuit breaker with per-user fairness
• Classic algorithm problems (trees, graphs, heaps) with real-time-systems twists (message ordering across shards, priority-based routing)

System Design Interview

One round, 60 minutes. Prompts are consumer-scale real-time:

• “Design the message-fanout system delivering messages across millions of channels to 200M users.”
• “Design the voice-routing system for 10K concurrent multi-user voice calls per region with optimal latency.”
• “Design the safety-moderation pipeline handling user reports and automated detection at scale.”
• “Design the Activities platform letting third-party apps run embedded in Discord channels.”

What works: real numbers (millions of concurrent WebSockets per gateway, specific latency budgets for message delivery, concrete shard / partition strategies), engagement with Discord’s publicly-shared architecture (ScyllaDB, Elixir messaging, Rust services), and consumer-scale operational reality. What doesn’t: generic web-app designs that don’t engage with real-time specifics.

Domain / Real-Time Deep-Dive

Role-specific. Sample topics:

Messaging / Elixir: walk through Erlang / OTP process model; discuss how GenServer supervision trees work; reason about a specific Elixir-at-scale challenge (hot-code loading, distributed clustering, process-limit tuning).

Rust services: discuss ownership / borrowing in a concurrent context; walk through async-Rust specifics (tokio runtime, async traits, futures composition); reason about why one would pick Rust over Go or Elixir for a specific problem.

Voice / media: discuss SFU vs MCU architecture; reason about WebRTC codec negotiation; describe server-side mixing vs client-side rendering trade-offs.

Moderation / ML: discuss balancing false-positive vs false-negative rates in abuse detection; reason about ML-model-deployment pipelines for CSAM classifiers; describe appeals-process design.

Product / User Round

Sample prompts:

• “What makes Discord great for communities? What’s broken?”
• “If you were improving Discord for server owners / moderators, what would you prioritize?”
• “How do you think about the tension between safety enforcement and community autonomy?”
• “What’s your take on Discord’s commerce / Activities platform direction?”

Candidates with real Discord usage (managing servers, participating in communities, using voice channels actively) have a meaningful edge. Those who use Discord only occasionally struggle.

Behavioral Interview

Key themes:

• Real-time systems experience: “Tell me about a high-scale real-time system you contributed to.”
• Consumer empathy: “Describe a time you deeply understood a consumer user’s problem.”
• Cross-functional work: “Discord’s work spans product, moderation, platform. Tell me about coordinating across multiple teams.”
• Ownership: “Describe a production incident you owned end-to-end.”

Preparation Strategy

Weeks 4-6 out: LeetCode medium/medium-hard in your team’s language. For Elixir / Rust / messaging teams, spend time reading the language ecosystem docs. Python and TypeScript are universally accepted for most rounds.

Weeks 2-4 out: read Discord’s engineering blog (famously good content). Posts on ScyllaDB migration, Rust rewrites, Elixir at scale, and voice infrastructure are all worth reading. For Elixir, Programming Elixir by Dave Thomas; for Rust at Discord scale, follow their public writing.

Weeks 1-2 out: use Discord deeply — manage a server, participate in real-time voice channels, try Activities. Form concrete opinions about the product. Mock system design with real-time-scale prompts.

Day before: review real-time-systems fundamentals; prepare product opinions; review behavioral stories.

Difficulty: 7.5/10

Solidly hard. Below Google L5 on pure algorithms; the real-time-systems specialty and consumer-scale depth are distinctive. Candidates with real-time / messaging / distributed-systems background have a clear edge. Strong generalists pass with focused prep.

Compensation (2025 data, US engineering roles)

• Software Engineer: $180k–$220k base, $130k–$250k equity (4 years), modest bonus. Total: ~$275k–$430k / year.
• Senior Software Engineer: $230k–$290k base, $280k–$520k equity. Total: ~$380k–$580k / year.
• Staff Engineer: $295k–$360k base, $550k–$1M equity. Total: ~$540k–$850k / year.

Private-company equity valued at 2021 peaks had to be adjusted downward post the broader tech reset; tender offers in 2024 reflected updated marks. 4-year vest with 1-year cliff. Compensation is competitive with mid-tier public tech; equity should be valued with meaningful illiquidity and valuation-volatility considerations.

Culture & Work Environment

Post-rightsizing, culture is more outcome-focused with tighter cost discipline than the 2020–2022 peak era. The engineering culture is genuinely technical with a gaming-adjacent sensibility — many engineers are active Discord users and community participants. Real-time systems expertise is respected; the company’s public engineering writing has attracted engineers specifically for the technical problems. Remote hiring is supported; SF HQ has meaningful in-office presence for specific teams. On-call matters for real-time infrastructure teams.

Things That Surprise People

• The engineering technical depth is genuinely distinctive. Real-time messaging and voice at Discord scale is novel engineering.
• The Rust / Elixir stack is real, not fashion. Performance-critical services are in Rust; messaging runs on Elixir / Erlang OTP.
• Trust & Safety is a substantial engineering investment, not an outsourced function.
• Product / user round genuinely weights authentic usage. Infrequent users struggle.

Red Flags to Watch

• Not using Discord. Interviewers will notice within minutes of the product round.
• Hand-waving on real-time systems. “We’d use WebSockets” isn’t enough; engage with scaling specifics.
• Weak Rust or Elixir knowledge when targeting teams that use them.
• Dismissing safety / moderation as boring. It’s one of Discord’s hardest engineering problems.

Tips for Success

• Use Discord actively. Run a server, join communities, try voice channels. Concrete experience signals.
• Read the engineering blog. Discord publishes depth that’s directly applicable to interview prep.
• Know Elixir / Rust fundamentals if targeting those teams. Surface familiarity isn’t enough.
• Engage with safety & moderation. Have an informed position.
• Prepare consumer-empathy stories. Discord’s users are real and varied — gamers, hobbyists, creators, educators, professionals.

Resources That Help

• Discord engineering blog (especially posts on ScyllaDB migration, Rust rewrites, Elixir messaging, voice infrastructure)
• Programming Elixir by Dave Thomas
• The Rust Programming Language (the official book)
• Designing Data-Intensive Applications (Kleppmann)
• Webrtc.org documentation for voice / video context
• Discord itself — active usage across different community types for 2+ weeks

Frequently Asked Questions

See also: Cloudflare Interview Guide • Zoom Interview Guide • Roblox Interview Guide