305 practice briefs
Build something you can check
Choose a mechanism, attempt the task, and compare what you made with the expected evidence. Hints stay separate from the prompt so you can try it first.
Search & IR
- Build a search tokenizer
- Build an inverted index
- Calculate a TF-IDF score by hand
- Calculate a BM25 score by hand
- Implement a BM25 ranker
- Compute precision and recall by hand
- Build a search eval harness
- Implement reciprocal rank fusion
- Cross-encoder rerank top-k
- Spell-correct a search query
- Browse vs search metrics
- Skip documents that cannot reach the top k
- NDCG@k
- Debias clicks by inverse propensity
Mathematics
- Compute a dot product and angle
- Multiply 2×2 matrices by hand
- Interpret an eigenvalue
- Apply Bayes' rule
- Expectation and variance of a discrete RV
- Summarize a small dataset
- Construct a 95% confidence interval
- Interpret a p-value correctly
- Fit a simple linear regression
- Compute a 2D gradient
- Independence and inclusion-exclusion
- Normal tail probability
- Correlation pitfalls
- Type I vs Type II errors
- Match distribution to scenario
- CLT and required sample size
- Standard error vs standard deviation
- Design an A/B test with power
- Interpret a sample ratio mismatch
- Beta-Binomial posterior update
- MLE for Bernoulli trials
- Compute cross-entropy in bits
- Compute covariance and correlation matrix
- Diagnose overfit with a train/holdout split
- Identify rank and column space
- PCA variance explained
- Log returns and annualized volatility
- Interpret an ACF plot
- Simulate a two-state Markov chain
- Sharpe ratio and max drawdown
- Design an honest momentum backtest
- Condition number and learning rate
- Latency against utilisation
- Distances concentrate as dimensions grow
- Count with inclusion-exclusion
- Log-sum-exp without overflow
Vector DB & ANN
- Embed a corpus and search it
- Implement similarity metrics
- Top-k with a bounded heap
- Build a brute-force vector index
- Explain HNSW search on paper
- Add metadata filtering to vector search
- IVF nprobe vs recall
- PQ codebook footprint
- Read a recall–latency curve
- Scalar quantise and rescore
- Cost a disk-resident search by I/O
- Tombstones erode recall
AI Systems
- Build a minimal RAG pipeline
- Build a structure-aware chunker
- Schema-constrained extraction
- Build a tool-calling loop
- Build an LLM eval
- Dot product and matmul shapes
- One SGD step on a quadratic
- Chain rule on a tiny graph
- Softmax and cross-entropy loss
- AdamW vs L2-on-gradient
- One BPE merge step
- Perplexity from average NLL
- Temperature scaling logits
- Sinusoidal position encoding
- Scaled dot-product attention weights
- Split embedding across heads
- Pre-LayerNorm residual block
- Overfit a single batch sanity check
- Checkpoint must include optimizer state
- GRPO group-relative advantage
- LoRA parameter count
- Estimate training token budget
- WASM linear memory growth
- WebGPU buffer upload size
- Perplexity vs classification accuracy
- Pack chunks into context budget
- Cap agent tool loop
- Route by cost and difficulty
- Diff two prompt versions
- Design exercise: Model Pre-training
- Design exercise: Model Quantization
- Design exercise: Open-Weight Models
Backend
- Implement a token-bucket rate limiter
- Implement idempotency keys
- Implement backoff with jitter
- Idempotent POST retry
- Hash and prefix API keys
- Verify webhook HMAC
- At-least-once job delivery
- Cache stampede
- Queue backpressure
- SLO error budget
- Design cursor pagination
- JWT revocation tradeoff
- Idempotent charge
- Design exercise: Observability & OpenTelemetry
- Exponential backoff with full jitter
- Parameterise instead of escaping
- Cursor pagination that survives inserts
Databases & Storage
- Build a memtable + SSTable flush
- Build a write-ahead log
- B-tree page split
- LSM compaction levels
- Object key layout
- Columnar vs row store
- Secondary index write cost
- Storage vs retrieval API
- Pick a shard key
- Read-your-writes under lag
- CAP under partition
- Design exercise: Transaction Processing
- Design exercise: Query Execution & Optimization
- Design exercise: Data Warehouses & Lakehouses
- Detect write skew under snapshot isolation
- Pick the join the planner should pick
- Spot the update anomaly
- Why LRU-K survives a scan
System Design
- Design a parking lot
- Design a vending machine
- Pluggable pricing strategies
- In-process pub/sub bus
- Notification factory
- Coffee-shop order decorator
- Concurrent bank transfers
- Undo / redo stack
- Hotel booking system
- L7 load balancer
- Distributed rate limiter
- Realtime chat system
- News feed (Twitter-style)
- Video streaming (Netflix-style)
- Ride-sharing dispatch
- Make three replicas converge
- Collaborative doc editor
- Search platform end-to-end
- Production RAG pipeline
- Consistent-hash key router
- Leader election walkthrough
- Realtime game server
- Design exercise: Real-time Application Engineering
- Size a write-heavy service
- Split functional from non-functional
- Token bucket rate limiter
- Pack and unpack a Snowflake id
- Decide the edge cache key
DSA & Implementation
- Two Sum with a hash map
- Valid palindrome with two pointers
- Longest substring without repeats
- Search in a rotated sorted array
- Binary tree level-order traversal
- Number of islands
- House robber (1D DP)
- Kth largest element with a heap
- Valid parentheses with a stack
- Reverse a linked list in place
- Trie prefix membership
- Generate all subsets
- Dijkstra shortest path
- Union-find component count
- Minimum path sum in a grid
- Activity selection greedy
- Merge overlapping intervals
- GCD and modulo arithmetic
- Find the single number (XOR)
- Two Sum
- Valid Palindrome
- Best Time to Buy and Sell Stock
- Contains Duplicate
- Maximum Subarray
- 3Sum
- Longest Substring Without Repeating Characters
- Valid Parentheses
- Merge Two Sorted Lists
- Binary Search
- Number of Islands
- Climbing Stairs
- Coin Change
- House Robber
- Product of Array Except Self
- Group Anagrams
- Top K Frequent Elements
- Invert Binary Tree
- Maximum Depth of Binary Tree
- Kth Largest Element in an Array
- Word Search
- Permutations
- Jump Game
- Merge Intervals
- Single Number
- Lowest Common Ancestor of a Binary Search Tree
- Serialize and Deserialize Binary Tree
- Course Schedule
- Rotting Oranges
- Task Scheduler
- Subtree of Another Tree
- Daily Temperatures
- Car Fleet
- Non-overlapping Intervals
- Sum of Two Integers
- Reverse Bits
- Amortised growth of a dynamic array
- Stable sort by one key
- Answer range-sum queries in O(1)
- Next greater element
- Rabin-Karp substring search
- Total weight of a minimum spanning tree
- Solve a divide-and-conquer recurrence
- k-th smallest by quickselect
Behavioral & Communication
Go-to-Market
Systems Foundations
- Concurrent bank transfers
- Design exercise: Operating System Mechanics
- Design exercise: Network Protocol Engineering
- Design exercise: Compute, Memory & Storage Hierarchy
- Design exercise: Runtime & Performance Engineering
- Design exercise: Security & Isolation Boundaries
- Encode a binary compatibility fixture
- Trace a program into process memory
Infrastructure & Platforms
- Implement backoff with jitter
- At-least-once job delivery
- SLO error budget
- Design exercise: Cloud Infrastructure
- Design exercise: Containers & Kubernetes
- Design exercise: CI/CD & Developer Environments
- Design exercise: Scheduling & Orchestration
- Design exercise: Reliability & Fault Tolerance
- Design exercise: Observability & OpenTelemetry
- Design exercise: Sandboxes & Execution Environments
- Design exercise: Infrastructure Automation
Distributed Systems
- Distributed rate limiter
- Realtime chat system
- Leader election walkthrough
- Cache stampede
- Queue backpressure
- Pick a shard key
- Read-your-writes under lag
- CAP under partition
- Design exercise: Event Streaming & Kafka
- Design exercise: Distributed Workflows & Temporal
- Design exercise: Distributed Failure Recovery
- Design exercise: Real-time Application Engineering
Inference & Serving
- WASM linear memory growth
- WebGPU buffer upload size
- Route by cost and difficulty
- Design exercise: vLLM & Inference Engines
- Design exercise: Continuous Batching
- Design exercise: KV Caching & PagedAttention
- Design exercise: FlashAttention & Attention Kernels
- Design exercise: Speculative Decoding
- Design exercise: GPU Utilization
- Design exercise: Inference Cost & Latency Optimization
- Design exercise: Local & On-device Inference
- Design exercise: Inference Hardware
Agent Systems
- Build a tool-calling loop
- Cap agent tool loop
- Design exercise: Agent Memory & Context Management
- Design exercise: MCP & Integrations
- Design exercise: Multi-agent Coordination
- Design exercise: Durable Agent Execution
- Design exercise: Agent Permissions & Sandboxing
- Design exercise: Browser & Computer-use Agents
- Design exercise: Agent Communication & Interfaces
- Design exercise: Long-running & Scheduled Agents
- Build a scoped repository instruction map
- Define an agent-readable environment contract
- Create a durable agent handoff
- Enforce runtime scope and escalation
- Separate maker claims from checker evidence
- Implement a bounded agent run lifecycle
- Evaluate and ablate an agent harness
Evaluation & AI Reliability
- Build an LLM eval
- Perplexity vs classification accuracy
- Diff two prompt versions
- Design exercise: Coding-agent Benchmarks
- Design exercise: Tool-use Evaluations
- Design exercise: AI Regression Testing
- Design exercise: Hallucination & Failure Detection
- Design exercise: Agent Observability
- Design exercise: Tracing & Replay
- Design exercise: Evidence-backed Verification
- Design exercise: Human Review Systems
- Design exercise: Quality, Cost & Latency Measurement
- Separate maker claims from checker evidence
- Evaluate and ablate an agent harness
Developer Tools & Code Intelligence
- Design exercise: Code Review Systems
- Design exercise: Static & Dynamic Analysis
- Design exercise: Testing Infrastructure
- Design exercise: Codebase Graphs
- Design exercise: Dependency & Blast-radius Analysis
- Design exercise: IDE & CLI Tooling
- Design exercise: Coding Agent Systems
- Design exercise: Repository Intelligence
- Design exercise: Software Supply-chain Health
- Design exercise: Automated Debugging & Remediation
- Build a scoped repository instruction map
- Define an agent-readable environment contract
- Enforce runtime scope and escalation
Application Engineering
Multimodal & Spatial Computing
All practice prompts
- Build a search tokenizer
- Build an inverted index
- Calculate a TF-IDF score by hand
- Calculate a BM25 score by hand
- Implement a BM25 ranker
- Compute precision and recall by hand
- Build a search eval harness
- Implement reciprocal rank fusion
- Embed a corpus and search it
- Implement similarity metrics
- Top-k with a bounded heap
- Build a brute-force vector index
- Explain HNSW search on paper
- Add metadata filtering to vector search
- Build a minimal RAG pipeline
- Build a structure-aware chunker
- Schema-constrained extraction
- Build a tool-calling loop
- Build an LLM eval
- Implement a token-bucket rate limiter
- Implement idempotency keys
- Implement backoff with jitter
- Build a memtable + SSTable flush
- Build a write-ahead log
- Two Sum with a hash map
- Valid palindrome with two pointers
- Longest substring without repeats
- Search in a rotated sorted array
- Binary tree level-order traversal
- Number of islands
- House robber (1D DP)
- Kth largest element with a heap
- Design a parking lot
- Design a vending machine
- Pluggable pricing strategies
- In-process pub/sub bus
- Notification factory
- Coffee-shop order decorator
- Concurrent bank transfers
- Undo / redo stack
- Hotel booking system
- L7 load balancer
- Distributed rate limiter
- Realtime chat system
- News feed (Twitter-style)
- Video streaming (Netflix-style)
- Ride-sharing dispatch
- Make three replicas converge
- Collaborative doc editor
- Search platform end-to-end
- Production RAG pipeline
- Consistent-hash key router
- Leader election walkthrough
- Realtime game server
- STAR: leadership story
- STAR: peer conflict
- STAR: tough decision
- STAR: cross-team collab
- STAR: a real failure
- STAR: explaining to a non-expert
- STAR: ruthless prioritization
- STAR: an unconventional idea
- STAR: acting on customer feedback
- STAR: end-to-end ownership
- Draft a product positioning statement
- Critique a real landing page
- On-page SEO audit
- Define a north-star metric
- Compute a dot product and angle
- Multiply 2×2 matrices by hand
- Interpret an eigenvalue
- Apply Bayes' rule
- Expectation and variance of a discrete RV
- Summarize a small dataset
- Construct a 95% confidence interval
- Interpret a p-value correctly
- Fit a simple linear regression
- Compute a 2D gradient
- Independence and inclusion-exclusion
- Normal tail probability
- Correlation pitfalls
- Type I vs Type II errors
- Match distribution to scenario
- CLT and required sample size
- Standard error vs standard deviation
- Design an A/B test with power
- Interpret a sample ratio mismatch
- Beta-Binomial posterior update
- MLE for Bernoulli trials
- Compute cross-entropy in bits
- Compute covariance and correlation matrix
- Diagnose overfit with a train/holdout split
- Identify rank and column space
- PCA variance explained
- Log returns and annualized volatility
- Interpret an ACF plot
- Simulate a two-state Markov chain
- Sharpe ratio and max drawdown
- Design an honest momentum backtest
- Dot product and matmul shapes
- One SGD step on a quadratic
- Chain rule on a tiny graph
- Softmax and cross-entropy loss
- AdamW vs L2-on-gradient
- One BPE merge step
- Perplexity from average NLL
- Temperature scaling logits
- Sinusoidal position encoding
- Scaled dot-product attention weights
- Split embedding across heads
- Pre-LayerNorm residual block
- Overfit a single batch sanity check
- Checkpoint must include optimizer state
- GRPO group-relative advantage
- LoRA parameter count
- Estimate training token budget
- WASM linear memory growth
- WebGPU buffer upload size
- Perplexity vs classification accuracy
- Valid parentheses with a stack
- Reverse a linked list in place
- Trie prefix membership
- Generate all subsets
- Dijkstra shortest path
- Union-find component count
- Minimum path sum in a grid
- Activity selection greedy
- Merge overlapping intervals
- GCD and modulo arithmetic
- Find the single number (XOR)
- Cross-encoder rerank top-k
- Spell-correct a search query
- Browse vs search metrics
- IVF nprobe vs recall
- PQ codebook footprint
- Read a recall–latency curve
- Pack chunks into context budget
- Cap agent tool loop
- Route by cost and difficulty
- Diff two prompt versions
- Idempotent POST retry
- Hash and prefix API keys
- Verify webhook HMAC
- At-least-once job delivery
- Cache stampede
- Queue backpressure
- SLO error budget
- Design cursor pagination
- JWT revocation tradeoff
- Idempotent charge
- B-tree page split
- LSM compaction levels
- Object key layout
- Columnar vs row store
- Secondary index write cost
- Storage vs retrieval API
- Pick a shard key
- Read-your-writes under lag
- CAP under partition
- Condition number and learning rate
- Two Sum
- Valid Palindrome
- Best Time to Buy and Sell Stock
- Contains Duplicate
- Maximum Subarray
- 3Sum
- Longest Substring Without Repeating Characters
- Valid Parentheses
- Merge Two Sorted Lists
- Binary Search
- Number of Islands
- Climbing Stairs
- Coin Change
- House Robber
- Product of Array Except Self
- Group Anagrams
- Top K Frequent Elements
- Invert Binary Tree
- Maximum Depth of Binary Tree
- Kth Largest Element in an Array
- Word Search
- Permutations
- Jump Game
- Merge Intervals
- Single Number
- Lowest Common Ancestor of a Binary Search Tree
- Serialize and Deserialize Binary Tree
- Course Schedule
- Rotting Oranges
- Task Scheduler
- Subtree of Another Tree
- Daily Temperatures
- Car Fleet
- Non-overlapping Intervals
- Sum of Two Integers
- Reverse Bits
- Design exercise: Operating System Mechanics
- Design exercise: Network Protocol Engineering
- Design exercise: Compute, Memory & Storage Hierarchy
- Design exercise: Runtime & Performance Engineering
- Design exercise: Security & Isolation Boundaries
- Design exercise: Cloud Infrastructure
- Design exercise: Containers & Kubernetes
- Design exercise: CI/CD & Developer Environments
- Design exercise: Scheduling & Orchestration
- Design exercise: Reliability & Fault Tolerance
- Design exercise: Observability & OpenTelemetry
- Design exercise: Sandboxes & Execution Environments
- Design exercise: Infrastructure Automation
- Design exercise: Event Streaming & Kafka
- Design exercise: Distributed Workflows & Temporal
- Design exercise: Distributed Failure Recovery
- Design exercise: Transaction Processing
- Design exercise: Query Execution & Optimization
- Design exercise: Data Warehouses & Lakehouses
- Design exercise: Model Pre-training
- Design exercise: Model Quantization
- Design exercise: Open-Weight Models
- Design exercise: vLLM & Inference Engines
- Design exercise: Continuous Batching
- Design exercise: KV Caching & PagedAttention
- Design exercise: FlashAttention & Attention Kernels
- Design exercise: Speculative Decoding
- Design exercise: GPU Utilization
- Design exercise: Inference Cost & Latency Optimization
- Design exercise: Local & On-device Inference
- Design exercise: Inference Hardware
- Design exercise: Agent Memory & Context Management
- Design exercise: MCP & Integrations
- Design exercise: Multi-agent Coordination
- Design exercise: Durable Agent Execution
- Design exercise: Agent Permissions & Sandboxing
- Design exercise: Browser & Computer-use Agents
- Design exercise: Agent Communication & Interfaces
- Design exercise: Long-running & Scheduled Agents
- Design exercise: Coding-agent Benchmarks
- Design exercise: Tool-use Evaluations
- Design exercise: AI Regression Testing
- Design exercise: Hallucination & Failure Detection
- Design exercise: Agent Observability
- Design exercise: Tracing & Replay
- Design exercise: Evidence-backed Verification
- Design exercise: Human Review Systems
- Design exercise: Quality, Cost & Latency Measurement
- Design exercise: Code Review Systems
- Design exercise: Static & Dynamic Analysis
- Design exercise: Testing Infrastructure
- Design exercise: Codebase Graphs
- Design exercise: Dependency & Blast-radius Analysis
- Design exercise: IDE & CLI Tooling
- Design exercise: Coding Agent Systems
- Design exercise: Repository Intelligence
- Design exercise: Software Supply-chain Health
- Design exercise: Automated Debugging & Remediation
- Design exercise: Web & Mobile Engineering
- Design exercise: UX & Interface Design
- Design exercise: Real-time Application Engineering
- Design exercise: 2D/3D Interactive Systems
- Design exercise: Multimodal Models
- Design exercise: Vision Models
- Design exercise: Pose & Motion Tracking
- Design exercise: Voice & Audio Systems
- Design exercise: Image & Video Generation
- Design exercise: Robotics Systems
- Design exercise: Spatial Interfaces
- Design exercise: Human-Computer Interaction
- Amortised growth of a dynamic array
- Stable sort by one key
- Size a write-heavy service
- Split functional from non-functional
- Answer range-sum queries in O(1)
- Next greater element
- Rabin-Karp substring search
- Total weight of a minimum spanning tree
- Detect write skew under snapshot isolation
- Pick the join the planner should pick
- Exponential backoff with full jitter
- Parameterise instead of escaping
- Latency against utilisation
- Distances concentrate as dimensions grow
- Skip documents that cannot reach the top k
- Token bucket rate limiter
- Pack and unpack a Snowflake id
- Solve a divide-and-conquer recurrence
- k-th smallest by quickselect
- Spot the update anomaly
- Why LRU-K survives a scan
- Cursor pagination that survives inserts
- Count with inclusion-exclusion
- Log-sum-exp without overflow
- NDCG@k
- Debias clicks by inverse propensity
- Scalar quantise and rescore
- Cost a disk-resident search by I/O
- Tombstones erode recall
- Decide the edge cache key
- Encode a binary compatibility fixture
- Trace a program into process memory
- Build a scoped repository instruction map
- Define an agent-readable environment contract
- Create a durable agent handoff
- Enforce runtime scope and escalation
- Separate maker claims from checker evidence
- Implement a bounded agent run lifecycle
- Evaluate and ablate an agent harness