Interview Prep

Real interview prep, organized by company and level

Mock rounds built from what candidates publicly report about the company you're interviewing at. Every question cites its source — Glassdoor, hellointerview.com, IGotAnOffer, 1Point3Acres. First one free, $9/mo after.

Run a free mock interview →How the mock works

~30 seconds
First one free
UK & US roles
Voice interview + report

Why per-company prep beats generic FAANG drills

A generic "FAANG interview prep" plan averages across companies whose loops are not actually similar. Google L4 has no system design round. Meta E4 always does. Amazon SDE2 fails coding-strong candidates on Bar Raiser if Leadership Principles don't surface. A practice loop built for the wrong company trains the wrong signals.

Each combo below practices against what is publicly documented for that specific company × level. Questions are sourced from public reports — Glassdoor, IGotAnOffer, hellointerview.com, Phil Shan's Medium, LeetCode Discuss, 1Point3Acres, Zhihu interview reports. No fabricated questions. No "we polled 10,000 candidates" stats. If a question is on the page, you can click through to the source thread it came from.

§ 01By cluster

How prep differs by company cluster

Tech-job clusters have meaningfully different filters even when they all 'test coding.' Pick the one closest to your target before locking in a study plan.

FAANG

Google · Meta · Amazon · Apple · Microsoft

FAANG comparison guide →

The most-studied cluster, but not interchangeable. Each runs a distinct rubric.

Google: Googliness + scale-aware reasoning
Meta: Move Fast + experiment-driven shipping with metric ownership
Amazon: 16 LPs scored continuously across the loop with Bar Raiser veto
Apple: team-specific (Photos vs Health vs Hardware run different loops)
Microsoft: growth mindset + customer impact + Azure-native stack signals

AI labs

OpenAI · Anthropic · Google DeepMind

AI labs guide →

Highest-paid cluster in tech with the highest filter density. PhD not required for general RE roles — depth equivalent to publication record substitutes.

OpenAI: production deployment + paper deep-dive rounds
Anthropic: values authenticity (non-technical values round is determinative)
DeepMind: 'The Quiz' — rapid-fire fundamentals on math/stats/RL/CS theory
Comp range: $300K (DeepMind UK) → $1.28M (OpenAI L6) per Levels.fyi

Quant firms

Citadel · Jane Street · Two Sigma · HRT · DE Shaw · Optiver

Quant firm comparison →

Highest absolute comp at senior. Math + probability rigor is the cross-firm baseline; firm-specific filters layer on top.

Citadel: commercial impact + pod alignment
Jane Street: least PhD-gated — Figgie game + Bayesian reasoning
Two Sigma: walk-forward backtest critique + sustainable culture
HRT: C++ depth + n→N proof skills
DE Shaw: pedigree-sensitive + adversarial PhD thesis defense
Optiver: 80-in-8 mental math filters ~50%+ of trader applicants

Fintech

Stripe · Coinbase · Plaid · Robinhood · Bloomberg

Different sub-domains but shared signals: idempotency, ledger-based design, regulatory awareness. Off-by-one in financial calc is hard-fail at all five.

Stripe: payments rails B2B integration depth
Coinbase: crypto-native architecture + custody
Plaid: bank-linking infrastructure + OAuth flows
Robinhood: consumer trading + order routing
Bloomberg: terminal + data + Code Review round

UK / European fintech

Wise · Monzo · Revolut

Distinct from US fintech. London-based with UK Skilled Worker visa pathway. Take-home formats dominate over whiteboard.

Wise: 'product engineer' philosophy + conversational system design
Monzo: skips whiteboard entirely — async take-home + Go-heavy backend
Revolut: 4-8h take-home with thread-safe SOLID + unit tests required

Tier-1 US unicorns

Stripe · Airbnb · Uber · Notion · Figma · Snap · Lyft

Mature private companies with established loops. Each has at least one signature round.

Airbnb: Host interviews by non-engineers + code review round
Uber: machine coding + Bar Raiser (modeled on Amazon)
Lyft: 90-min Laptop Programming Test (open-internet, IDE-based)
Notion: ~4h interview block with culture-fit weight
Figma: Web frontend + C++ + WebAssembly knowledge expected
Snap: Kind/Smart/Creative scored across every round

Big-tech adjacent

TikTok · ByteDance · Snowflake · Databricks · Cloudflare · Palantir

Each has at least one distinctive round-format quirk that catches generic preppers off-guard.

TikTok: algorithm-heavy (LeetCode-medium, 2-3 problems timed)
ByteDance HQ: 15-min Hard DP at HM round + Mandarin requirement
Snowflake: 2h combined DSA + system design phone screen
Databricks: dedicated concurrency / multithreading round + VP veto
Cloudflare: 'Orange Cloud' 30-min strict behavioral + AI-assisted coding
Palantir: Decomposition + Re-engineering rounds + 12-month rejection cooldown

How Applr's mocks differ from generic LeetCode-style

Most mock interview tools train one signal: can you produce a working solution under time pressure. That's necessary but not sufficient. The candidates who actually convert offers also signal correct trade-off reasoning, behavioral fit per company values, and ability to articulate decisions out loud — and those signals are what most candidates fail on.

Applr loads what is publicly documented about the company you're targeting before each session: which behavioral framework gets scored (LP / Googliness / Move Fast / Kind-Smart-Creative / etc.), what stack signals matter (Go for Uber, PHP/Hack for Meta, C++ for HRT/Optiver, JAX for DeepMind), what failure modes candidates report getting flagged on (Bar Raiser claim drift, Apple secrecy violation, Cloudflare corporate-speak). Sessions are bilingual (EN + ZH) — useful for international students who want to first-draft in native language before delivering in English.

Practice the loop you are actually facing, against what candidates say that loop is like.

§ 02Available now