/ QUESTION BANK
Filter by track, sub-track, difficulty and company. Click any question to start a mock.
How would you implement a Promise polyfill?
How would you implement Promise.all() from scratch?
Why doesn't async/await automatically run code in parallel?
What's the difference between Promise.all(), Promise.race(), Promise.any(), and Promise.allSettled()?
What does .then() return?
Why does an async function always return a Promise?
What's the difference between microtasks and macrotasks?
Why does Promise.then() run before setTimeout(..., 0)?
Why is class just syntactic sugar over prototypes?
Why can't arrow functions be used as constructors?
How would you implement an LRU Cache?
What happens during the Creation Phase of the Execution Context?
How does this work in default, implicit, explicit, and new binding?
How many closures are created in nested functions?
Why does typeof sometimes throw a ReferenceError?
How does JavaScript find a variable? (Lexical Scope and the Scope Chain)
How would you deep clone an object with circular references?
Explain how you would benchmark and compare the performance of two vector DB solutions (e.g., Pinecone vs. Milvus) for a real-time image similarity service, including dataset preparation, query patterns, and cost analysis.
How would you design a scalable vector similarity search pipeline that handles 10M+ high-dimensional embeddings per day while guaranteeing sub-second latency for top-k retrieval?
In a multi-tenant SaaS product, what strategies would you employ to isolate and secure each tenant's vector index while still sharing underlying hardware resources?
Explain how you would combine approximate nearest neighbor (ANN) algorithms (e.g., HNSW, IVF-PQ) with a fallback exact search to meet both speed and recall requirements in a recommendation system.
Describe a production-grade data ingestion workflow that continuously updates embeddings from a streaming source (Kafka) into a vector DB without causing index corruption or downtime.
How can you implement dynamic re-ranking that merges traditional BM25 scores with dense vector similarity scores in real time, and what trade-offs does this introduce?
Discuss the impact of dimensionality reduction techniques (e.g., PCA, UMAP) on index size and query accuracy when storing billions of vectors — when would you apply them, and how would you validate the loss of information?
What monitoring metrics and alerting thresholds would you set up to detect vector drift or degradation in similarity quality over time?
Propose a cost-effective strategy to shard a large vector index across multiple geographic regions while preserving global nearest-neighbor semantics.
How would you integrate role-based access control (RBAC) and field-level encryption into a vector database that stores sensitive user embeddings?
Outline a CI/CD pipeline for deploying schema migrations (e.g., adding new metadata fields) to a live vector index without interrupting ongoing search queries.
In a hybrid search scenario, how would you architect a system that simultaneously supports keyword, hybrid, and pure vector queries while maintaining a single API surface?
How do you implement a supervisor agent in LangGraph?
What are the main multi-agent patterns in LangGraph?
How do you implement a tool that modifies graph state directly?
How do you bind tools to an LLM in LangGraph?
What is create_react_agent and when should you use it vs building a custom graph?
How do you implement parallel tool execution in LangGraph?
How do you handle tool errors in LangGraph?
What is a ToolNode and how does it work?
What is the difference between interrupt_before and interrupt_after?
How do you resume a graph after a human-in-the-loop interrupt?
What is the interrupt() function and how does it differ from compile-time interrupts?
How do you implement human-in-the-loop in LangGraph?
How do you implement cross-thread memory (shared state across conversations)?
What is checkpoint_ns and when does it matter?
How do you update state manually between graph invocations?
How does time-travel work in LangGraph?
Explain the concept of checkpoints, threads, and runs in LangGraph.
What is a thread_id and how does it relate to checkpointing?
What checkpointer backends does LangGraph support?
What is a checkpointer in LangGraph and why is it critical?
What is a conditional edge and how do you implement one?