What you will learn
- Select 3 projects that tell a coherent portfolio story
- Apply the Range-Depth-Recency framework to project selection
- Structure each project's GitHub README and presentation for maximum recruiter impact
Why this matters
Freshers either have too few projects (one college assignment) or too many weak ones (10 tutorial repos). Neither works. The strongest portfolios have exactly 3 projects that show range, depth, and recent activity. Three is enough to demonstrate breadth; few enough that each one is polished. Hiring managers at Indian product companies consistently say they would rather see 3 excellent projects than 8 mediocre ones.
The Range-Depth-Recency framework
Your 3 projects should cover three axes:
1. Range — show you can work across domains
Do not make all 3 projects the same type. Pick from different categories:
| Slot | Purpose | Example types |
|---|---|---|
| Project 1 | Your strongest, most complete project | Full-stack app, AI pipeline, CLI tool |
| Project 2 | Shows a different skill | If P1 is frontend-heavy, P2 should be backend/infra/AI |
| Project 3 | Shows initiative or specialization | Open-source contribution, research tool, dev tool, domain-specific app |
2. Depth — at least one project should go beyond the obvious
"Depth" means you went past the tutorial:
- You wrote tests
- You handled edge cases and errors
- You optimized performance
- You wrote documentation
- You made a non-obvious architecture decision
3. Recency — at least one project should have commits from the last 3 months
Stale profiles signal that you stopped learning. Keep at least one project active.
Common portfolio archetypes for Indian CS students in 2026
The "AI + Full-Stack" portfolio (most versatile)
- Full-stack web app — React/Next.js + Python/Node backend + PostgreSQL. Shows you can build a complete product.
- AI/ML project — RAG pipeline, fine-tuned classifier, or LLM-powered tool. Shows you understand the AI stack.
- Open-source contribution or developer tool — CLI tool, VS Code extension, or meaningful PR to an established project. Shows you can work in existing codebases.
The "Backend + Systems" portfolio (for infrastructure roles)
- API-heavy backend — multi-tenant system, payment integration, real-time features.
- Systems project — custom cache, job queue, or distributed service with performance benchmarks.
- DevOps/tooling — CI/CD pipeline, monitoring setup, or infrastructure-as-code project.
The "Full-Stack Specialist" portfolio (for product companies)
- Complete SaaS app — auth, billing, multi-tenancy, deployed and live.
- UI-rich project — dashboard, data visualization, or complex form-heavy app showing frontend depth.
- Side project with real users — even 10 users signals product thinking.
The project README template
Every portfolio project needs a README that a recruiter can scan in 60 seconds:
# Project Name
One-sentence description of what this does and why.
## Demo
[Link to live demo] or [Screenshot/GIF]
## What I Built
2-3 sentences explaining the project. Focus on the problem
it solves, not just the tech used.
## Architecture
- **Frontend:** React, Tailwind CSS
- **Backend:** FastAPI, PostgreSQL
- **AI/ML:** sentence-transformers, FAISS
- **Infra:** Docker, GitHub Actions CI/CD
[Optional: architecture diagram]
## Key Technical Decisions
1. **Chose PostgreSQL over MongoDB** because the data is
relational (users, orders, products with foreign keys).
2. **Used FAISS locally instead of Pinecone** to keep the
project free and self-contained for reviewers.
3. **Implemented rate limiting at the API gateway level**
rather than per-endpoint to reduce middleware complexity.
## How to Run
\`\`\`bash
git clone https://github.com/you/project.git
cd project
cp .env.example .env # fill in your API keys
docker compose up # or: npm install && npm run dev
\`\`\`
## Test
\`\`\`bash
npm test # or: pytest
\`\`\`What to cut from your portfolio
Remove these from your pinned/featured repos:
- Tutorial follow-alongs — even if you modified them slightly
- College assignments — unless yours went significantly beyond the requirements
- Half-finished projects — a broken demo is worse than no demo
- "Awesome list" forks — forking a curated list is not a project
- Projects with no README — if you cannot be bothered to explain it, why should the recruiter bother to open it
Presenting your portfolio in interviews
When asked "Walk me through a project," use this 2-minute structure:
- Problem (15s): What problem does this solve? Why did you build it?
- Architecture (30s): High-level system design. Draw it if there is a whiteboard.
- Hardest part (45s): The most technically challenging decision or problem you solved.
- Result (15s): What works, what you measured, what you would do differently.
- Invite questions (15s): "Happy to go deeper on any part — the retrieval pipeline was the most interesting piece."
## Portfolio selection worksheet
List your projects and score each 1-5:
| Project | Completeness | Technical Depth | Uniqueness | Recency | Total |
|---------|-------------|-----------------|------------|---------|-------|
| | /5 | /5 | /5 | /5 | /20 |
| | /5 | /5 | /5 | /5 | /20 |
| | /5 | /5 | /5 | /5 | /20 |
| | /5 | /5 | /5 | /5 | /20 |
| | /5 | /5 | /5 | /5 | /20 |
Pick your top 3 by total score.
Then check: do they show range (different tech/domain)?
If two are too similar, swap the lower-scoring one for a
different-category project even if its total score is lower.Try it yourself
- List every project you have built (including incomplete ones)
- Score each project using the worksheet above
- Select your top 3 using the Range-Depth-Recency framework
- For each of the 3 projects, write or update the README using the template
- Pin these 3 on your GitHub profile and verify each has:
- A clear one-sentence description
- A "How to Run" section that actually works (test it from a fresh clone)
- At least one screenshot or demo link
Common mistakes
- Three projects in the same domain — three React CRUD apps show you know one thing, not three. Diversify across backend, AI, frontend, or DevOps.
- Listing a project with zero tests — even 5 unit tests signal that you understand testing. Projects with zero tests signal tutorial-quality work.
- README says "work in progress" — if it is not finished, do not feature it. Finish it or replace it with something complete.
- No "How to Run" instructions — if a recruiter clones your repo and cannot run it in 5 minutes, they will not try harder. Test your setup instructions from scratch.
- Identical commit messages — "initial commit," "update," "fix" repeated 20 times suggests you do not care about craft. Clean up the history or at least use meaningful messages going forward.
Quick quiz
Q: A student has 5 projects: (1) React to-do app, (2) React e-commerce app, (3) RAG pipeline, (4) Express REST API, (5) contributed a PR to an open-source CLI tool. Which 3 should they pin?
How to explain it in an interview
"My portfolio has three projects chosen for range: a full-stack multi-tenant SaaS app showing I can build complete products, a RAG pipeline showing AI engineering skills, and a PR I contributed to an open-source project showing I can work in existing codebases. Each has a clean README with setup instructions, architecture notes, and my key technical decisions."
Checklist
- Selected 3 projects using the Range-Depth-Recency framework
- Each project has an updated README following the template
- All 3 projects can be cloned and run from scratch (tested it yourself)
- The 3 projects are pinned on your GitHub profile
- At least one project has commits from the last 3 months
- Can explain each project in 2 minutes using the Problem-Architecture-Hardest Part-Result structure