Full-stack is the most over-claimed title on Indian IT CVs. We have screened CVs claiming full-stack with React, Node, PostgreSQL, MongoDB, AWS, Kubernetes, Terraform, GraphQL, WebSockets, and six other buzzwords, all in under four years of experience. Ninety percent of the time those CVs belong to someone who has touched a React component separately from a Node endpoint and has never shipped a feature that connects the two. The honest definition of full-stack is simple: someone who can ship a complete feature without handoff. Schema, API, UI, tests, deploy. One person, one pull request, end-to-end. The interview protocol below separates them in about 40 minutes using one concrete question — "walk me through the last feature you shipped end-to-end, starting with the database." True full-stack is worth roughly 30 percent more than either pure front-end or pure back-end at the same level. Get the taxonomy right at intake.
Two or more years building production React, Vue, or Angular UIs. Comfortable with state management, async data fetching, form handling, accessibility basics, responsive CSS. Not "I added a button to a page." Ask them to whiteboard the component tree for a feature they built.
Two or more years writing API endpoints in Node, Python, Go, or Java. Understands auth (JWT, sessions, OAuth), request validation, error handling with meaningful status codes, database queries beyond SELECT *, basic caching. Has shipped at least one endpoint handling money, user data, or a third-party integration.
Can design a relational schema for a CRUD-plus feature under interview conditions. Knows when to normalize and when to denormalize for read performance. Understands indexes well enough to add one and explain why. Has written at least one migration they regretted.
Has shipped at least one non-trivial feature where they personally wrote schema, API, UI, tests, and participated in deploy. Side projects count if live with real users. A capstone or Udemy project with three users does not count.
Knows how their code reaches production. Can describe the deploy platform (Vercel, Netlify, AWS ECS, Kubernetes, plain EC2), the CI pipeline, and at least one rollback. Developers who say "DevOps handles that" are not full-stack, they are mid-stack.
Modern full-stack is heavily TypeScript on both ends. Strong asset for Bangalore product companies and Western client teams. Ask if they have shipped both a TypeScript frontend and type-safe backend.
React Native, Flutter, or true native. Broadens the "full" in full-stack, especially valuable for Indian D2C and fintech clients where the mobile app is the primary surface.
Can set up a basic GitHub Actions or GitLab CI pipeline themselves. Avoids the "I cannot deploy without DevOps" bottleneck common in mid-level full-stack engineers.
Knows when to add an index, lazy-load components, paginate, cache. Beyond "make it work." Ask about the last performance issue they fixed — they should name the tool they used.
Recognizes bad UI, even if they cannot design from scratch. Can implement from Figma without pixel-level dictation. Rare and valuable on small teams with no dedicated designer.
Walk me through the last feature you shipped end-to-end. Start with the database schema and end with the deploy. Use your most recent production feature.
What to listen for
Specific tables and columns, specific endpoints with HTTP methods, specific UI components and state management decisions, specific deploy platform and rollback. Vague "I used Next.js and deployed to Vercel" without decisions means they worked on someone elses full-stack app. Strong candidates mention trade-offs they considered and rejected.
How do you decide what state lives on the front-end versus the back-end? Give me a specific example from something you built.
What to listen for
Source-of-truth thinking. UI-only state (modal open, dropdown selection) on the client. Domain state (user profile, order status) on the server with client caching. Not "everything goes in Redux." Strong answers reference React Query, SWR, or tRPC.
A page in your app is loading slowly. Users complain it takes 6 seconds to render. Walk me through your investigation, step by step.
What to listen for
Network tab first for API response times, then DevTools performance panel for render cost, then back-end query analysis (slow query log, APM tool), then decide where the bottleneck is before optimizing. Weak: "I would add Redis caching." That is a solution, not an investigation.
How do you handle authentication in a feature you are building today, from login through protected API calls? Walk me through the full flow.
What to listen for
Specific approach — JWT with short access tokens and refresh tokens, session cookies with CSRF, OAuth with a provider. Awareness of pitfalls — where to store tokens, expiry and refresh, CSRF, logout on all devices. Not "I use Auth0" without understanding what Auth0 is doing.
Tell me about a design decision you made on a feature that turned out wrong. What did you do when you discovered it?
What to listen for
Honest example, learning extracted, refactor or revert. "I stored user preferences in a JSON blob and regretted it when we needed to query by individual preference. We migrated to a normalized table over two sprints."
Design a comment system for a blog post. Start with the schema, then the API, then how the UI would render it.
What to listen for
Normalization (post, comment, user tables), threading (parent_comment_id with recursion or closure table), soft deletes, moderation fields, pagination (cursor vs offset), UI considerations for nested rendering. Strong candidates ask clarifying questions before diving in.
You deployed a feature yesterday. This morning your inbox has five bug reports. Walk me through the first hour.
What to listen for
Triage by impact first — does anyone lose data or money? Reproduce locally with the most critical bug. Decide between hotfix and full rollback based on scope. Communicate timeline. Strong candidates check logs and error-tracking (Sentry, Rollbar) in the first 10 minutes.
What does "full-stack" mean to you, and where do you personally have the most gaps?
What to listen for
Self-aware answer. Nobody is equally strong across database, back-end, front-end, mobile, and infra. Strong candidates name their weak area — "I can build React UIs but my CSS is average, I lean on Tailwind." Weak candidates claim equal strength everywhere, which is almost always a lie.
Score each candidate against these weighted criteria. Total: 100%.
| Criterion | Weight | Signal |
|---|---|---|
| Front-end depth | 25% | Shipped non-trivial React, Vue, or Angular UIs with state management, async data, accessibility. Can whiteboard a component tree. Has opinions about CSS strategy. |
| Back-end depth | 25% | Shipped real endpoints with auth, input validation, error handling, database queries beyond basic CRUD. Has dealt with a real production back-end issue. |
| End-to-end ownership | 20% | Shipped full features themselves, not handed off mid-stack. Can describe the entire path from schema to deploy for a specific feature. |
| Database fluency | 15% | Designs relational schemas confidently. Knows query basics beyond SELECT *. Has added an index to fix a real performance problem. |
| Operational awareness | 15% | Deploys what they built. Has debugged a real production incident. Reads their own error logs and metrics in the first hour after a deploy. |
Claims full-stack but cannot draw a basic schema for a blog-post-with-comments feature in under 10 minutes
Only contributed to existing front-ends or back-ends — has never started a project from an empty repository
Does not know how their code reaches production — deploy pipeline is a black box owned by DevOps
Lists eight or more frameworks in skills section with no specific project proving depth in any
Cannot describe a real feature shipped end-to-end in the last 6 months — either unemployed or not actually shipping
Upload Full-Stack Developer CVs and let AI score every candidate against the same 42-point evidence rubric.
Try CVPRO Free