What C4GT Actually Is, and How to Get Into DMP 2027
I'm a second year B.Tech student. Right now I'm rebuilding the frontend of India's national 104 health helpline as a C4GT DMP Fellow. Six months ago I didn't know this program existed.
This is the guide I wanted then. No fluff, no motivational framing. Just what it is and how to get in.
1. What is C4GT
Code for GovTech connects developers to open source software that governments and public institutions actually run. Not simulations of civic tech. Live systems: health, education, welfare delivery, identity. These serve millions of people and are maintained in public repositories.
Why this matters more than it sounds: most government software in India is now built as Digital Public Goods, meaning open source, openly licensed, and reusable across states. That means the code is on GitHub and you can contribute to it. A platform serving a hundred million citizens will take your pull request if it's good.
The gap C4GT fills is that these repos are hard to enter cold. Large, sparsely documented, deep domain context. C4GT provides the structure: a fellowship, a mentor from the organization that owns the codebase, and a scoped project you can actually finish.
2. What is DMP
DMP is the fellowship track. It's a paid, mentored, fixed term program where you work on one Digital Public Good project with a real mentor from the maintaining organization. Details: pl-app.iiit.ac.in/c4gt.
How it runs, in shape:
- Organizations publish project ideas with scope and required skills
- You pick a project and apply through a form on the C4GT site
- The form runs in stages, asking for your repositories, your education, your background, and why this project
- Resume and proposal uploads are offered but optional
- Selected fellows are paired with a mentor from that organization
- Midterm evaluation at the halfway point: you present shipped work
- Final evaluation at the end
- Stipend paid across milestones
Two things that surprise people. First, it's not an internship where you're handed tickets. You shape the plan with your mentor and own delivery. Second, the code goes into the real repository. My PRs are merged into a system that live call operators use today.
3. What kind of projects
A real example: mine.
India's 104 health helpline is what a citizen dials for free health advice, medical guidance, and grievance reporting. Behind that number sits AMRIT, maintained by Piramal Swasthya. A health advisory officer answers, and inside one application does caller identification, beneficiary lookup, symptom capture, clinical decision support, and case closure.
That application was built on Angular 4, a framework long past end of life. npm audit on it reports 183 vulnerabilities. My project: rebuild the frontend on Angular 20 with standalone components, zoneless change detection, signals, ZardUI, Tailwind v4, SNOMED CT clinical coding, and CDSS integration.

At midterm: 30 PRs merged, 131 commits, 52 of 60 end to end scenarios passing.
Other projects run across health, education, welfare, agriculture, and identity, spanning backend, frontend, data, mobile, and DevOps. Browse the current list on the C4GT community platform; scope and stack vary widely.
4. How to apply
- Read the project list at pl-app.iiit.ac.in/c4gt. Filter to projects whose stack you can already work in. You'll be learning the domain, so don't also be learning the language.
- Join the C4GT Discord. This is the actual center of gravity. Announcements, mentor Q&A, other applicants. Join before applications open, not after.
- Find the repo and read it. Every project maps to a public GitHub repo. Clone it. Build it. Get it running locally. This alone puts you ahead of most applicants.
- Open a small PR before you apply. Fix a typo, tighten error handling, improve a README. A mentor who has already merged something of yours is evaluating a known quantity.
- Ask your questions in public. Discord or a GitHub issue, not DMs. Mentors notice who engages carefully.
- Fill in the application form on the C4GT site. You choose your project first, then work through several stages of questions covering your repositories, your education, and your background. Resume and proposal uploads are optional. Watch the deadline. There's no extension.
5. What to prepare
GitHub. Pinned repos with real READMEs: what it does, how to run it, a screenshot. Ten abandoned tutorial clones read worse than two finished projects. Commit history matters more than star count.
Resume, optional. One page. Lead with projects and what they do, not coursework. Link every repo. Skip the skills percentage bars.
The form answers. This is where it's actually decided. There is no formal proposal to write. You pick a project, then work through stages of questions inside the form, and those answers are what get read. What to have ready before you start:
- Your repository links, strongest one first
- Why this project, in terms of the problem it solves rather than how excited you are
- What you would build, concretely: components, endpoints, screens
- Evidence you read the code, referencing actual files and actual functions
- Your education and background, which the form asks for directly
That fourth point is the differentiator. Most applicants describe the project back to the organization in more words. If your answer says "I looked at the auth service in src/app/core and here's how I'd migrate it," you're in a different pile.
The resume and proposal uploads are there if you want them. Use them only if you have something the form fields can't hold, like a migration plan or a longer technical note. Skip them if all they would do is repeat your answers.
6. What it's actually like
Honest version, three months in.
The codebase is nothing like coursework. Sixteen years of accumulated decisions, most undocumented. Functions that look wrong and aren't. They're compensating for backend quirks nobody wrote down. I learned this by deleting a few and breaking things. Reading legacy code well is a skill nobody teaches and everybody needs.
Senior devs will be direct with you. My mentor rejects PRs that aren't right. That felt sharp for about a week and then became the most useful thing in the fellowship. Nobody is protecting your feelings, and that's the respect.
AI assisted development helps in a specific, limited way. I used Claude Code for implementation and CodeRabbit for PR review. It genuinely collapsed the time to understand an unfamiliar codebase, and CodeRabbit caught real null safety and error handling bugs before review. But my hardest problem was not solvable by asking: integrating the CZentrix telephony system, which turned out to be an iframe communicating over postMessage and documented nowhere. That took instrumenting the old app and reading actual traffic. AI is fast at what code does and unreliable at why. Verify everything against the running system.
Write things down daily. I keep notes on every confirmed API contract, every environment quirk, every bug and cause. By month two I was answering my own questions from my own notes constantly. Nothing else I did had that return.
The moment that made it real: a live inbound call arrived on the test system. Caller's number populated. Timer started. Workspace loaded. A real phone call from somewhere in India, rendering into a UI I'd written. That's the difference between a project and infrastructure.
7. Links
- C4GT: https://codeforgoodtech.in/
- C4GT on GitHub: https://github.com/Code4GovTech/C4GT
- Apply / community platform: https://pl-app.iiit.ac.in/c4gt
- Discord: https://discord.gg/codeforgovtech
- AMRIT / PSMRI repos: https://github.com/PSMRI
- My GitHub: https://github.com/Aarti-panchal01
- My portfolio: https://aarti-tech-portfolio.vercel.app
- LinkedIn: https://linkedin.com/in/aartipanchal01
8. Start now
Applications for DMP 2027 aren't open yet, which is exactly why this is the right time.
Do this today: join the Discord. Pick one project from last cycle's list. Clone the repo and get it running. When applications open you'll have months of context while everyone else is reading the project description for the first time.
I applied assuming I wasn't ready. Second year, no industry experience, no idea if my answers were any good. They were good because I'd read the code.
That's the whole trick. Go read the code.
*Aarti Panchal. C4GT 2026 DMP Fellow at Piramal Swasthya, B.Tech AIML '28, PES University Bengaluru.
Questions welcome on My LinkedIn
Top comments (0)