← practice

Lead Engineer Focus — E5

meta

A session built entirely around the current Meta senior-IC (E5) loop — the level some orgs call “lead.” The round breakdown, then 58 of the highest-frequency coding questions with Java solutions, plus system-design and behavioral prompts.

Sourcing note: exact live frequency is Premium-gated and NDA-limited. This is a compiled best-view from recent public E5 guides and reports — a strong signal, not a leaked list.

The loop

Coding × 2

core bar
Two ~35-minute sessions in CoderPad, typically two mediums each (hards increasingly common). One round is now often the AI-enabled pilot — you may use an assistant, so the bar shifts to how you direct and verify it. Dynamic programming is effectively banned; expect strings/parsing, trees, graphs, intervals, and hashing.

System / Product Architecture

level-setting
Meta pushes for concrete numbers — throughput, storage math, cache invalidation — not hand-waving. Product-architecture variants center a user-facing feature (a widget, a feed) and drill the API + data model.

Behavioral ("Jedi")

can down-level
A standalone 45-minute session on conflict, ambiguity, failure, and cross-functional work. Now weighted enough to move an E5 to E4 on its own — prepare specific, structured stories.

Team matching

unscored
Not scored, but it feeds the final decision and your starting team. Come with informed questions about the org and the problem space.

Coding — 58 current high-frequency questions

No dynamic programming (effectively banned at Meta). Strings/parsing, trees, graphs, intervals, and hashing dominate.

Strings & Parsing

Meta's most-loved bucket

Arrays, Intervals & Search

prefix sums, sweeps, binary search

Trees & BST

BFS columns, DFS state up/down

Graphs, Heap & Design

flood fill, topo sort, heaps, design

System / product architecture

Meta pushes for concrete numbers — throughput, storage, cache invalidation. Practice driving the API and data model, not staying high-level.

Design an Ad Click Aggregator

Streaming ingestion, windowed counts, exactly-once vs. approximate — throughput math is expected.

Design an Online Game Leaderboard

Top-K at scale, rank queries, hot-key sharding, real-time updates vs. periodic snapshots.

Design LeetCode / a Judging System

Submission queue, sandboxed execution, result fan-out, and idempotent retries.

Design a Ticket Booking System

Seat reservation, holds/expiry, and avoiding double-booking under contention.

Design a Top-K / Trending System

Count-min sketch vs. exact counts, decay windows, and read-heavy fan-out.

Product Architecture: Top-K Songs Widget

Client + API + data model for a user-facing feature; drill the read path and caching.

Product Architecture: Price-Drop Tracker

Watches, change detection, and notification fan-out with dedup.

Behavioral (“Jedi”)

A standalone 45-minute round, now heavy enough to down-level. Prepare specific, structured stories (situation → action → measurable result).

Resources — articles & guides

Curated external prep. Some (LeetCode's Meta tag, a couple of guides) need a free account.