Top Tech Interview Questions & Answers for IT Placements

React Server Components execute strictly on the server during the build or request time, shipping zero client-side JavaScript bundle for their logic. They can directly query databases, read backend filesystems, and protect sensitive API keys. Client Components ('use client') run on the browser, manage interactive state (useState, useReducer), and bind browser event listeners (onClick, onChange).

Node.js uses an event-driven, single-threaded architecture powered by libuv. The event loop traverses phases: Timers (setTimeout, setInterval), Pending Callbacks, Poll (retrieving new I/O events and executing their callbacks), Check (setImmediate), and Close Callbacks. Microtasks (process.nextTick and resolved Promise callbacks) execute immediately after each phase before moving to the next.

Key strategies include: 1) Creating appropriate Single-Field or Compound Indexes matching query predicates. 2) Utilizing the explain('executionStats') method to verify index usage and eliminate collection scans (COLLSCAN). 3) Employing projection to return only essential fields instead of entire documents. 4) Leveraging aggregation pipelines with indexed $match stages placed early. 5) Using connection pooling and caching frequent reads via Redis.

Session-based authentication is stateful: the server stores user session data in memory or a database (like Redis) and sends a session ID cookie to the browser. JWT (JSON Web Token) is stateless: the token itself contains digitally signed cryptographic claims (payload) encoded in Base64Url. The server verifies the signature with a secret/private key without querying a session store, making JWT highly scalable across microservices.

RAG connects a Large Language Model (LLM) to an external vector database containing proprietary knowledge. When a query is made, relevant context documents are semantically retrieved via embeddings and injected into the LLM prompt. RAG eliminates model hallucination, requires no costly GPU training, allows real-time knowledge updates, and provides traceable source citations, unlike fine-tuning which alters model weights without guaranteeing factual recall.

Overfitting occurs when a model learns training data noise, yielding high training accuracy but poor validation performance. It is detected by comparing training loss against validation loss curves. Solutions include: 1) Regularization (L1 Lasso, L2 Ridge, Dropout in neural nets). 2) Pruning decision trees or reducing model complexity. 3) K-Fold Cross-Validation. 4) Early stopping during iterative gradient descent. 5) Increasing training dataset volume through data augmentation.

Supervised Learning trains on labeled datasets to map inputs to known outputs (e.g., Regression, Classification). Unsupervised Learning discovers hidden patterns, groupings, or clusters from unlabeled data without predefined ground truths (e.g., K-Means, PCA). Reinforcement Learning involves an autonomous agent interacting with an environment, learning an optimal policy through trial-and-error rewards and penalties (e.g., Q-learning, PPO in robotics/gaming).

A GROUP BY clause collapses multiple rows into a single summary row per group (e.g., SUM(salary) per department). Window functions (such as ROW_NUMBER(), RANK(), DENSE_RANK(), and SUM() OVER (PARTITION BY ...)) calculate aggregated or relative metrics across a defined set of rows while preserving each individual row in the output table.

The primary mitigation is using Parameterized Queries (Prepared Statements) or Object-Relational Mappers (ORMs) where user input is treated strictly as data parameters rather than executable SQL syntax. Additional defenses include implementing stored procedures, enforcing principle of least privilege on database users, validating and allowlisting input types, and deploying a Web Application Firewall (WAF).

Symmetric encryption uses a single shared secret key for both encryption and decryption (e.g., AES-256, ChaCha20); it is computationally fast and ideal for encrypting large volumes of data. Asymmetric encryption uses a mathematically linked key pair: a public key for encryption and a private key for decryption (e.g., RSA, ECC); it is used for secure key exchanges (TLS/HTTPS handshakes) and digital signatures.

Virtual Machines virtualize physical hardware through a hypervisor, requiring a full guest operating system (kernel, binaries, libraries) for each VM, consuming substantial memory and CPU. Docker containers virtualize at the OS level, sharing the host machine's Linux kernel while running isolated user-space processes. This makes containers lightweight, spinning up in milliseconds with minimal resource overhead.

An Amazon Virtual Private Cloud (VPC) is a logically isolated virtual network in the AWS cloud. Security Groups operate at the individual EC2 instance level, are stateful (inbound allowed traffic automatically permits return outbound traffic), and support allow rules only. Network Access Control Lists (NACLs) operate at the subnet level, are stateless (inbound and outbound rules evaluated separately), and support both allow and deny rules.

Structure your answer with the STAR framework (Situation, Task, Action, Result): 1) State the business problem your project solves. 2) Detail your technical stack and architecture decisions (e.g., why Next.js over vanilla React). 3) Highlight the biggest technical obstacle you overcame (e.g., database concurrency, latency optimization). 4) Share verifiable outcomes: live deployment URL, test coverage, or benchmark performance metrics.

Explain a systematic debugging methodology: 1) Reproduce the bug in an isolated test environment. 2) Inspect error stack traces, server logs, and network payloads using browser DevTools or APM tools. 3) Formulate hypotheses and test them iteratively with minimal reproducing code. 4) Research authoritative documentation, official GitHub issues, and RFC specifications. 5) When needed, ask team peers concisely with clear problem context, attempted solutions, and minimal reproduction steps.

Tricity (Mohali IT City, QuarkCity, Chandigarh IT Park) is one of North India's fastest-growing enterprise and startup software ecosystems, home to 450+ product companies, global IT consultancies, and innovative engineering firms. Starting here provides intensive hands-on ownership on real production systems, close mentorship under senior engineers, and accelerated technical career growth.