DEV Community: Chaitanya Kulthe

India’s Unemployable Engineers: When Degrees Teach 1990s Tech in a AI World

Chaitanya Kulthe — Fri, 18 Apr 2025 19:04:31 +0000

So there I was, chatting with a BCA second‑year student over chai. I asked, “How’s college these days? What topic did your professor cover today?” He grinned and said, “Sir spent a full 30 minutes explaining the <table> tag in HTML!” He said it with such pride—as if mastering tables was the pinnacle of web development. Inside, I felt a knot in my stomach: how could our education system still waste precious time on stuff any developer learns in ten minutes on YouTube?

That tiny moment sums up a much bigger problem. On one hand, India’s tech industry is booming—projected to hit $300 billion by FY 26. On the other, only 42.6 percent of our graduates are deemed employable, and barely 10 percent land relevant jobs straight out of college. We’re churning out bodies, not builders.

In countless labs, professors hawk dusty slides on HTML 4.0, Java Swing, or fire up Turbo C to spend a full lecture walking students through writing a simple fraction‑calculator in C—tasks any developer today could spin up in minutes with modern languages and tools. Ask a simple question—“Sir, can we see how to deploy this API on AWS?”—and you’re told to “focus on your syllabus.” It’s not arrogance; it’s insecurity. Admitting you don’t know the latest tech feels like losing face.

Meanwhile, fresh graduates walk straight into corporate bootcamps only to be weeded out again. Infosys recently let go of 240 trainees who flunked basic foundation tests, despite extra coaching and mock exams. At TCS, recruiters admit they hire from only 10–15 percent of colleges because of this massive skill gap. And a NASSCOM report warns that 40 percent of working IT pros will need reskilling over the next five years just to keep their jobs.

So what is real education? To me, it means:

Facing the Unknown: Being dropped into a nasty bug or a broken pipeline and figuring it out yourself.
Embracing Failure: Running into errors at 2 AM and celebrating the “Aha!” when it finally works.
Learning Out Loud: Push that messy, half-working project to GitHub. It’s not about perfection—it’s about progress, reflection, and leveling up in public..

That’s how India’s tech legends got started. The founders of Infosys, Wipro, Flipkart, Zomato—they didn’t wait for college to teach them Docker or TensorFlow. They built things on the side, joined coding clubs, read blogs at midnight, and hustled for real experience.

At this year’s Startup Mahakumbh, Commerce Minister Piyush Goyal slapped us with a challenge: “While China churns out semiconductors and trains its own AI models, too many Indian startups are content delivering pizza or groceries”. He’s right—and the blame falls partly on our classrooms. You can’t build an AI chip when your degree is stuck on table tags and syntax trees.

Students, don’t let your syllabus cap your potential. Fire up freeCodeCamp or NPTEL, build a REST API, containerize it, deploy it on a cloud free tier, and blog every step. Host your code on GitHub. That living portfolio shouts “I can ship code” far louder than a perfect CGPA ever will.

Teachers, this is part love letter, part reality check.
I know I don’t have the right to tell you how to teach—you’ve seen way more than I have. But maybe… just maybe, it’s time for a shift.
Call industry folks for guest talks—your slides from 2003 aren’t cutting it anymore.
Swap “write a for loop” exams for real builds: a live web app, a Dockerized microservice, or a chatbot that doesn’t just spit “Hello World.”
Update your syllabus every 3–4 years, not every retirement cycle.
And please—if you don’t know something, don’t fake it. Google it. Learn with your students. That humility? That’s leadership.
It’s 2025, not 1995. Let’s not prepare students for a tech world that doesn’t exist anymore.

Institutions and policymakers, carve out Centres of Excellence and fund labs—yes—but also tie accreditation to real outcomes: graduate employability, industry partnerships, and live project work. Budget 2025’s ₹500 crore for AI in Education is a start, but unless that money reaches lecture halls and lab benches, it’s just another headline.

A college degree should be a launchpad, not a cage. If your institution won’t evolve, you must. Because in 2025, the future belongs to builders, innovators, and lifelong learners—those who refuse to let a dusty syllabus dictate their worth. Be the spark that drags our education system, kicking and screaming, into the modern age.

Your degree expires. Your curiosity doesn’t. Keep building. 🔥

China’s Manus Agent by Monica: What You Actually Need to Know

Chaitanya Kulthe — Sun, 09 Mar 2025 22:05:25 +0000

Introduction

Silicon Valley isn’t the only player in AI anymore. China just dropped Manus Agent, a new AI tool built by a company called Monica, and it’s designed to do things Western tools like ChatGPT can’t. But who’s behind it? Why should you care? And what does it mean for countries like India? Let’s cut through the hype.

Who is Monica?

Monica isn’t a person—it’s a Chinese tech firm that’s been quietly working on AI for sectors like healthcare, logistics, and agriculture. Unlike flashy startups, Monica focuses on practical solutions for governments and industries. Think of them as the “backstage crew” of China’s tech revolution. Their goal? To build AI that works in the real world, not just in labs.

Manus Agent: The Basics

Manus Agent isn’t here to write poems or crack jokes. It’s built to solve messy, real-world problems. Here’s what makes it different:

Trained on Asian Data:
- Most AI (like OpenAI) learns from Western-centric data—books, websites, and social media dominated by English. Manus studied Chinese, Hindi, Bahasa, and regional dialects. This means it understands phrases like India’s “jugaad” or China’s “guanxi” better than ChatGPT ever could.
Works Without the Internet:
- Imagine using AI in an Indian village with spotty 4G. Manus can run on local servers or even a smartphone. That’s huge for farmers, factories, or hospitals in areas with poor connectivity.
Follows China’s Rules:
- It’s programmed to comply with China’s strict laws on data privacy (which prioritize national security over individual privacy). For example, it might refuse to answer questions critical of the Chinese government.

Why Should India Care?

India’s tech scene is booming, but most AI tools here rely on Western platforms. Manus could change that.

Opportunities:
- Indian startups could license Manus for cheap, hyper-local AI solutions—like predicting crop yields in Punjab or managing traffic in Mumbai.
- Its offline feature is perfect for rural India, where internet access is still a luxury.
Risks:
- Data Sovereignty: If an Indian company uses Manus, where does your data go? China’s laws let the government access data for “security” reasons.
- Competition: India’s own AI projects (like those by Tata or Infosys) might struggle to match Manus’ funding and scale.

Manus vs. OpenAI: No, It’s Not a Replacement

OpenAI (ChatGPT): Your go-to for brainstorming, writing, or casual chat. It’s creative but generic.
Manus: Like a specialist doctor. It won’t write your resume, but it will diagnose a rare disease using local medical data.

Example:

If a Chennai hospital uses Manus, it could analyze Tamil-language patient records to spot dengue outbreaks. ChatGPT? It’d need perfect English inputs.

Privacy and Control

Manus’ biggest strength is also its weakness. By design, it aligns with China’s vision of AI—efficient, obedient, and state-approved.

For Indian Users:
- Pros: Affordable, localized, works offline.
- Cons: Your data might feed into China’s AI ecosystem. If you’re a small business, is that a risk worth taking?

The Future: AI Nationalism

Manus isn’t just a tool—it’s part of China’s plan to lead AI globally. Countries like India now face a choice:

Adopt foreign tools (like Manus or OpenAI) for quick fixes.
Invest in homegrown AI tailored to local needs (but this takes time and money).

For India, this could mean more partnerships (like Reliance teaming up with Monica) or stricter rules to keep data within borders.

Final Word

Manus Agent isn’t “better” than other AI tools like ChatGPT—it’s simply built for different challenges. Its focus on diverse, local data and operation in low-connectivity settings could be a game changer for many regions. However, with its strengths come trade-offs, especially regarding data privacy and regulatory control. Ultimately, whether tools like Manus Agent will help bridge the technology gap or deepen reliance on foreign systems depends on the choices made by businesses and governments around the world.

🚀 GitHub Copilot’s Latest Update: Your AI Pair Programmer Just Got Smarter!

Chaitanya Kulthe — Fri, 07 Feb 2025 08:10:29 +0000

Hey folks! If you’re a developer, you’re going to love this—GitHub just dropped some serious upgrades to Copilot, making it feel less like a code suggestion tool and more like an actual coding assistant. The highlight? Agent Mode, which takes Copilot from a helpful sidekick to something that actively fixes, debugs, and improves your code on the fly. Let’s break it down in plain English.

🤖 Meet Agent Mode: Copilot, But Smarter

We all love Copilot for its code completions, but imagine if it could debug itself, suggest fixes, and even explain runtime errors. That’s exactly what Agent Mode brings to the table.

What can it do?

✅ Detect mistakes in its own suggestions

✅ Fix errors automatically before you even realize there’s a bug

✅ Suggest terminal commands based on your workflow

✅ Analyze and resolve runtime issues (yep, self-healing code!)

It’s like having a junior dev sitting beside you, spotting errors, and saying, "Hey, I think I messed up there. Let me fix it!" instead of just dumping a chunk of code and leaving you to figure it out.

How do you enable it?

To try out Agent Mode:

Download VS Code Insiders
Enable the Agent Mode setting for GitHub Copilot Chat
In the Copilot Edits panel, switch from Edit to Agent

And boom! Your Copilot just leveled up.

🛠 Copilot Edits: Now Available for Everyone

If you’ve been using the preview version of Copilot Edits, good news—it’s now officially available in VS Code! This feature lets you make changes across multiple files at once, using just natural language.

Imagine typing:

"Refactor this function to improve performance and update the related tests."

And watching Copilot take care of everything. It even provides explanations so you can review and tweak as needed. No more bouncing between files and manually making changes.

🌟 Project Padawan: The Future of AI-Driven Code Contribution

Now, this is where things get really interesting. GitHub teased a new experimental project called Padawan, which could turn Copilot into a full-fledged software engineering agent.

🚀 What does that mean?

You’ll soon be able to assign issues directly to Copilot
It will generate fully tested pull requests
It’ll even work with human reviewers and iterate based on feedback

Imagine GitHub Copilot as a real open-source contributor—this could be a game-changer for teams working on large-scale projects.

🎯 Final Thoughts: Copilot is Becoming More Than Just an AI Assistant

This latest update makes one thing clear—GitHub Copilot is no longer just a fancy autocomplete tool. It’s evolving into an AI-powered dev assistant that can code, debug, and even collaborate like a real team member.

If you haven’t tried these features yet, now’s the perfect time to dive in and see how they can boost your workflow. What do you think? Could Agent Mode or Project Padawan change the way we code? Let’s discuss in the comments! 👇

🔗 Watch the full breakdown: GitHub Copilot: The Agent Awakens

Happy coding! 🚀

What is WebRTC and How Does It Work?

Chaitanya Kulthe — Wed, 09 Oct 2024 07:14:25 +0000

WebRTC (Web Real-Time Communication) is a cool technology that helps us talk, share videos, or even send messages directly between two people using a web browser. You don’t need to install anything! It’s already built into browsers like Chrome, Firefox, and others.

Let’s break down how WebRTC works:

1. MediaStream:

This is like a magic door that lets your web browser grab audio or video from your microphone or camera. Imagine you want to video call your friend. The browser opens up the camera and microphone so you can hear and see each other. 🎥🎤

2. RTCPeerConnection:

This is the part that connects two people directly—just like passing a note directly to your friend in class! Instead of passing through a server or a third person, the connection goes straight from you to your friend. WebRTC does its best to create this direct link (peer-to-peer connection). 📨

3. RTCDataChannel:

Want to send messages or files to your friend? This is where RTCDataChannel comes in. It’s like a little conveyor belt that sends data quickly and securely between both people. 🛤️

What Does WebRTC Do?

It works across different platforms like phones, tablets, and laptops.
It’s smart—it can change how it works to give you the best video and audio quality, even if your internet is slow.

What Happens Behind the Scenes?

To help the two people (or devices) talk directly to each other, WebRTC uses a few more magic tricks. These tricks help your devices find the best path to send messages, even if there are obstacles like firewalls (imagine walls that stop unwanted things from coming in).

STUN Server (Session Traversal Utilities for NAT):

A STUN server is like a guide. It tells your device, “Hey! This is how you look to the outside world!” When your device asks, “Who am I?” the STUN server responds with your public IP address (this is like your home address on the internet). With this information, your device can find the best way to talk to the other device directly.

TURN Server (Traversal Using Relays around NAT):

Sometimes, your device can’t connect to the other device directly (maybe there’s a very strong firewall). This is where a TURN server helps! The TURN server is like a middleman—it takes your message, holds it for a bit, then passes it along to the other device. It's like sending a letter through the post office if you can't deliver it yourself. 📮

However, using a TURN server might slow things down a bit, but it’s still super helpful when a direct connection is not possible.

ICE (Interactive Connectivity Establishment):

Before your devices can start talking, they need to figure out the best path to send data, even if they are behind firewalls or routers. ICE is like a map that helps your devices find the best way to connect.

The Secret Ingredient: SDP (Session Description Protocol)

Whenever your devices are getting ready to connect, they exchange some important information called SDP. SDP is like an invitation card that tells each device what kind of video, audio, or data to expect and how to set it up. This makes sure both sides are on the same page. 📜

How Does All This Work Together?

Starting the Conversation:
- You (Peer A) want to call your friend (Peer B).
- You both send your public IP addresses (home addresses) using the STUN server to figure out how to connect directly.
If Direct Connection Fails:
- If the connection can’t happen directly (because of firewalls or other issues), the TURN server helps by relaying the messages.
Sending Messages or Video:
- Once everything is set, you both start sharing audio, video, or text messages through a data stream directly or through the TURN server.

Why WebRTC is Awesome

No plugins needed: Everything happens inside your web browser.
Adaptable: WebRTC knows how to give you the best experience even with bad internet.
Free: You don’t have to pay to use WebRTC, and it works on most devices.

In Simple Words:

WebRTC helps two people (or devices) talk or share videos easily, without needing extra apps. It’s like a shortcut for communication, and if things get tricky, it has backup plans to keep things working smoothly.