# Intermediate Interview Preparation Guide (2-4 Years) Source: https://www.techinterview.org/intermediate-interview-guide/ Updated: 2026-07-03 · techinterview.org Intermediate Interview Preparation Guide *For candidates with 2-4 years of software engineering experience* ## Overview As an intermediate engineer, you're expected to tackle more complex algorithmic problems and demonstrate deeper understanding of system design fundamentals. This 6-8 week intensive guide prepares you for interviews at top tech companies including FAANG. ## Prerequisites - Comfortable with all basic data structures and algorithms - Can solve 80%+ of LeetCode Easy problems - Understanding of [Big O notation](/big-o-cheat-sheet/) - Experience with at least one major programming language - Professional software development experience ## Timeline: 6-8 Weeks ### Week 1: Advanced Arrays and Strings **Focus:** Complex Patterns and Optimization **Topics:** - [Sliding window](/post/3233474160/coding-interview-two-pointers-sliding-window-patterns-array-string-problems-fast-slow-pointer-variable-window/) variations — Know both the fixed-size window and the variable-size window that shrinks from the left when a constraint breaks. Interviewers usually push you from a brute-force O(n²) scan to a single O(n) pass. - Two pointers advanced techniques — Practice opposite-end pointers on sorted arrays and same-direction pointers for in-place work. The common trap is off-by-one errors when you skip past duplicates. - String matching algorithms (KMP, Rabin-Karp) — Understand how KMP's prefix table avoids re-scanning and how Rabin-Karp uses a rolling hash. You rarely code these from scratch, but explaining why they beat naive O(n·m) matching does come up. - Subarray problems — Prefix sums plus a hash map turn "count subarrays that sum to k" from O(n²) into O(n). Watch for negative numbers, which break the sliding-window assumption. **Target Problems (15-20):** - Longest Substring Without Repeating Characters (Medium) — The canonical variable window: expand the right edge, and jump the left pointer past the last seen index of a repeat. - Minimum Window Substring (Hard) — Track character counts with a map and a have/need counter so you know exactly when the window is valid before shrinking it. - Longest Palindromic Substring (Medium) — Expand around each center for O(n²); mention Manacher's O(n) algorithm if asked to optimize. - Find All Anagrams in String (Medium) — A fixed-size window plus a frequency count that you slide one character at a time. - Sliding Window Maximum (Hard) — A monotonic decreasing deque holds the candidate maxima and gives O(n) overall. ### Week 2: Dynamic Programming Fundamentals **Focus:** 1D and 2D DP **Topics:** - Memoization vs tabulation — Memoization is top-down recursion with a cache; tabulation builds a table bottom-up. Be ready to convert one to the other and to explain the space savings when you only need the last row or two. - State definition and transitions — Most DP interviews are won or lost here: name what dp[i] means in one sentence, then write the recurrence. If you can't define the state cleanly, the code won't come. - 1D DP (House Robber, Climbing Stairs) — These reduce to dp[i] depending on dp[i-1] and dp[i-2], so you can drop the array down to two variables. - 2D DP (Unique Paths, Edit Distance) — Grid problems where dp[i][j] depends on the cells above and to the left; fill the table by hand before you code it. **Target Problems (10-15):** - Longest Increasing Subsequence (Medium) — The O(n²) DP is the baseline; the O(n log n) version with binary search is the follow-up interviewers ask for. - Coin Change (Medium) — An unbounded-knapsack shape where dp[amount] is the minimum coins; returning -1 for an unreachable amount is the edge case people miss. - Word Break (Medium) — dp[i] is true if some split point j has dp[j] true and s[j:i] in the dictionary; a set lookup keeps it fast. - Longest Common Subsequence (Medium) — The 2D template that Edit Distance and many string DPs build on: match characters to go diagonal, otherwise take the best of up or left. - Edit Distance (Medium) — Three transitions (insert, delete, replace); walking through "horse" → "ros" on the whiteboard shows you understand it. ### Week 3: Advanced Tree and Graph Algorithms **Focus:** Complex Traversals and Path Finding **Topics:** - Binary tree serialization/deserialization — Pick one traversal (preorder with null markers is simplest) and use it for both directions. Be explicit about how you encode nulls and delimit values. - Lowest Common Ancestor — For a BST, use the ordering to walk down; for a general binary tree, recurse and return the node where the two targets split. - Graph traversals (DFS, BFS) — Know when BFS gives shortest paths on an unweighted graph and how to track visited nodes to avoid cycles. Be able to write both iteratively and recursively. - [Topological sort](/post/3233461698/topological-sort-interview-patterns/) — Use Kahn's algorithm (BFS on in-degrees) or DFS post-order; both detect cycles, which is usually the hidden requirement. - [Union-Find](/post/3233474181/coding-interview-union-find-disjoint-set-deep-dive-path-compression-union-by-rank-connected-components-kruskal-accounts-merge/) (Disjoint Set) — Path compression plus union by rank gives near-constant operations. It's the fast answer for connectivity and cycle detection in undirected graphs. **Target Problems (15-20):** - Binary Tree Maximum Path Sum (Hard) — Each node returns the best single-branch gain upward while updating a global max that may split through the node. - Serialize and Deserialize Binary Tree (Hard) — Preorder with explicit null markers, rebuilt with an index or queue on the way back. - Course Schedule (Medium) — A topological-sort and cycle-detection problem in disguise; "can you finish" means "is the graph acyclic". - Number of Islands (Medium) — Flood fill with DFS or BFS while marking visited cells; the follow-up is often counting with union-find. - Clone Graph (Medium) — A hash map from original node to copy prevents infinite loops while you DFS or BFS the graph. ### Week 4: Advanced Dynamic Programming **Focus:** Knapsack, Subsequences, and Optimization **Topics:** - 0/1 Knapsack and variations — Each item is take-or-skip, with dp over items and capacity. Recognizing a problem as knapsack in disguise (subset sum, partition) is the real skill. - Unbounded Knapsack — Item reuse is allowed, so you iterate capacity forward; Coin Change is the classic instance. - DP on strings — Palindromes, subsequences, and matching all share a 2D dp[i][j] over two indices or two strings; get comfortable choosing the axes. - DP with bitmasks — Use an integer as a set when n is roughly 20 or fewer, as in traveling-salesman-style problems, where dp[mask][i] tracks the visited set and current position. **Target Problems (10-15):** - Partition Equal Subset Sum (Medium) — Reduces to subset-sum for half the total; if the total is odd, answer no immediately. - Target Sum (Medium) — Assigning + and - signs becomes a subset-sum count once you rearrange the equation. - Palindrome Partitioning II (Hard) — Precompute an is-palindrome table, then run a 1D DP for the minimum cuts. - Burst Balloons (Hard) — Interval DP: think about the last balloon to burst in each range, not the first. - Regular Expression Matching (Hard) — 2D DP where "*" either drops the preceding element or consumes one more character; the star transitions trip most people up. ### Week 5: Heaps, Tries, and Advanced Data Structures **Focus:** Specialized Structures **Topics:** - Min/Max Heap operations — Push and pop are O(log n) and building a heap is O(n). Most languages hand you a priority queue, so practice inverting the comparator to get a max-heap. - Trie implementation — A tree of children maps with an end-of-word flag; it turns prefix search and autocomplete into O(word length). - LRU Cache design — A hash map plus a doubly linked list gives O(1) get and put; be ready to explain why a plain array or a single map isn't enough. - Design problems — These test data-structure composition and clean APIs more than raw algorithms. Talk through each operation and its target complexity before you write code. **Target Problems (10-15):** - Implement Trie (Medium) — Insert, search, and startsWith over a children map, the base you extend for word-search and autocomplete follow-ups. - LRU Cache (Medium) — The map-plus-linked-list pattern; the follow-up is often LFU, which adds frequency buckets. - Top K Frequent Elements (Medium) — A count map then a heap of size k, or bucket sort by frequency for O(n). - Find Median from Data Stream (Hard) — Two heaps, a max-heap for the lower half and a min-heap for the upper half, kept balanced so the median sits at the tops. - Design Twitter (Medium) — Merge each followed user's recent tweets with a heap; it's the news-feed fan-out problem in miniature. ### Week 6: System Design Fundamentals **Focus:** Scalability and Architecture **Topics:** - Load balancing — Know round-robin, least-connections, and consistent hashing, and where each fits. Interviewers ask how you keep a single load balancer from becoming the bottleneck. - Caching strategies — Compare cache-aside, write-through, and write-back, and explain eviction (LRU) and TTLs. Be ready to discuss cache invalidation and stampede protection. - [Database sharding](/post/3233459955/database-sharding/) — Split data by a shard key (hash or range) to scale writes; the hard parts are hot keys, cross-shard joins, and resharding. - CAP theorem — Under a network partition you choose consistency or availability. Name real systems on each side and tie the choice back to the product's needs. - Microservices basics — Understand service boundaries, inter-service communication, and the operational cost they add. Interviewers probe for when a monolith is the better call. **Practice Designs:** - URL Shortener (bit.ly) — Focus on the encoding for short codes, collision handling, and read-heavy caching. Capacity estimates for the keyspace size come up often. - [Rate Limiter](/post/3233474159/system-design-rate-limiter-token-bucket-sliding-window-leaky-bucket-distributed-rate-limiting-api-gateway/) — Compare token bucket, leaky bucket, and sliding-window counters, and decide where the counters live in a distributed setup. - Key-Value Store — Talk through partitioning, replication, and consistency; this is where consistent hashing and quorum reads and writes show up. - Notification System — Design fan-out to email, SMS, and push with queues, retries, and de-duplication. Delivery guarantees and user preferences are the depth points. ### Weeks 7-8: Mock Interviews and Polish **Focus:** Interview Performance **Activities:** - 10-15 mock coding interviews — Do these live with a person or a timer, and treat thinking out loud as part of the practice, not an afterthought. - 3-5 system design mock interviews — Record yourself or get feedback on structure; most people lose points on time management, not on knowledge. - [Behavioral interview practice](/category/interview-prep/) — Rehearse your stories until the structure is automatic, then work on trimming each one to about two minutes. - Review [all patterns](/algorithm-patterns-cheat-sheet/) and problem types — Re-solve one problem per pattern from memory to confirm you can recall the approach, not just recognize it. ## Study Resources ### Books: - **Cracking the Coding Interview** (all chapters) - **System Design Interview** by Alex Xu (Volume 1) - **Designing Data-Intensive Applications** by Martin Kleppmann ### Online Platforms: - **LeetCode:** Focus on Medium problems, aim for 150-200 total - **AlgoExpert:** Comprehensive video explanations - **Educative:** Grokking the Coding Interview, Grokking System Design ## Interview Expectations by Company ### Google (L3/L4): - 2-3 coding rounds (45-60 min each) - Focus: Algorithms, data structures, complexity analysis - Expect medium-hard problems - 1 round may include system design discussion ### Amazon (SDE2): - 2-3 coding rounds - 1-2 rounds focused on [Leadership Principles](/post/3233474665/amazon-leadership-principles-cheat-sheet/) - Expect medium problems with edge cases - System design for senior positions ### Facebook/Meta (E4): - 2 coding rounds (medium-hard) - 1 system design round - 1 behavioral round (culture fit) - Focus on problem-solving speed ### Microsoft (60/61): - 2-3 coding rounds - Mix of easy-medium and medium-hard - Strong emphasis on testing and edge cases - System design discussion ## Problem-Solving Patterns to Master - **Sliding Window:** Substring, subarray problems - **Two Pointers:** Pair problems, palindromes - **Fast & Slow Pointers:** Cycle detection - **Merge Intervals:** Scheduling, overlaps - **Cyclic Sort:** Missing numbers in range - **In-place Reversal of Linked List:** List manipulation - **BFS/DFS:** Tree and graph traversal - **Top K Elements:** Heaps - **Binary Search:** Sorted arrays, search space - **Backtracking:** Permutations, combinations - **Dynamic Programming:** Optimization problems ## System Design Preparation ### Core Concepts to Master: - Horizontal vs vertical scaling — Adding more machines versus buying a bigger one; know the ceiling on vertical scaling and why stateless services scale out more easily. - Load balancers (L4 vs L7) — L4 routes on IP and port, L7 on HTTP content; L7 lets you route by path or header at some extra processing cost. - Caching (Redis, Memcached) — Redis adds data structures and optional persistence; Memcached is a simpler in-memory cache. Be ready to place caches at the client, CDN, or service layer. - Databases ([SQL vs NoSQL](/post/3233459967/sql-vs-nosql/), [indexing](/post/3233461821/database-indexing-interview-guide/), sharding) — Justify the choice by access pattern and consistency needs, not preference, and know how an index speeds reads while slowing writes. - Message queues (Kafka, RabbitMQ) — Kafka is a durable, replayable log for streaming; RabbitMQ is a broker for task queues. Both decouple producers from consumers and absorb traffic spikes. - CDNs and edge caching — Serve static assets close to users to cut latency; know cache-control headers and how invalidation works. - Microservices architecture — Independent deploys and clear boundaries at the cost of distributed-systems complexity; mention observability and data consistency across services. ### Design Interview Template: - **Requirements (5 min):** Clarify functional and non-functional requirements - **Estimation (3 min):** Users, QPS, storage needs - **High-Level Design (10 min):** Draw main components - **Deep Dive (20 min):** Focus on 2-3 components in detail - **Bottlenecks (5 min):** Identify and resolve ## Behavioral Interview Prep **Prepare 8-10 stories using [STAR method](/post/3233460379/behavioral-interview-questions-2026-star-method-amazon-leadership-principles-and-winning-answers/):** - **Situation:** Context and background - **Task:** What needed to be done - **Action:** What you specifically did - **Result:** Outcome and learnings **Cover these themes:** - Leadership and influence — Show impact without formal authority: a time you drove a decision or brought a team along. Ownership is what senior bars look for here. - Conflict resolution — Pick a disagreement with a peer or manager and focus on how you reached alignment, not on who was right. - Technical challenges overcome — Choose a problem with real ambiguity and walk through your reasoning and tradeoffs, not just the final fix. - Failed project and learnings — Own a genuine failure and be specific about what you changed afterward; a sanitized "I just work too hard" answer falls flat. - Innovation and creativity — Have an example where you challenged the default approach and it paid off, with a measurable result if you can attach one. ## Success Metrics **You're ready when:** - ✓ Can solve 60%+ of LeetCode Medium problems - ✓ Comfortable with 20%+ of Hard problems - ✓ Can design 3-4 common systems from scratch - ✓ Pass 70%+ of mock interviews - ✓ Can explain tradeoffs clearly **Target Statistics:** - 150-200 LeetCode problems solved - 5-10 system designs practiced - 10+ mock interviews completed ## Final Tips - Focus on understanding patterns, not memorizing solutions - Practice explaining your thought process out loud - Time yourself consistently (45 min for coding, 45 min for system design) - Don't ignore behavioral prep—it matters at senior levels - Apply broadly—interviewing is a skill that improves with practice **Remember:** At this level, companies assess not just your coding ability, but also your communication, system design thinking, and cultural fit. Prepare holistically for the best results.