AI for India’s School Innovators: How Mr Parth Devariya Demonstrated Application Building, Vehicle Telematics, and Data Sovereignty at PM SHRI RMS Parul University on Day 2

Most AI workshops at the school level demonstrate what AI can do. The Day 2 session at PM SHRI RMS demonstrated what students can build with AI in under ten minutes and what they should know about where their data goes when they do.

Mr. Parth Devariya, AI and Technology Consultant at GFuture Tech Pvt. Ltd., opened the Day 2 AI session on 29 April 2026 at the Hospital Auditorium with a deceptively simple question: what can we do with AI? The audience of PM SHRI School, Jawahar Navodaya Vidyalaya, and Kendriya Vidyalaya participants responded with the expected answers (finding solutions, creating music, generating images, building presentations, augmented reality, virtual reality). Mr. Devariya’s role was to take the answers beyond the obvious into the operational. The session operated as a structured part of the Three-Day Regional Mentoring Session jointly convened by AICTE, the Ministry of Education’s Innovation Cell, the Wadhwani Foundation, and PIERC at Parul University.

The entire 20-member jury panel comprised Dr. Arvind Deshmukh, Ms. Anbumathi M, Mr. Parth Devariya, Mr. Hardik Kharva (Centre Head, VSS, PIERC), Ms. Sonal Sudani (Incubation Manager, PIERC), and Mr. Umang Panchal (Assistant Professor, PIET), Mr. Anup Chaudhary (Incubation Manager), Mr. Umang Panchal, Mrs. Sonal Sudani (PIERC), Mr. Hardik Kharwa, Ms. Sujaya Bhattacharjee, Mr. Himansu Das, Ms. Vanshika Muchhara, Dr. Partkumar Sapariya, Dr. Bhavin Dhanavade, Dr. Prashant Khanna, Dr. Sneha Soni, Dr. Saurabh Parmar, Ms. Kajol Patel, Mr. Vivek Joshi, Ms. Riddhi Mehta, and Mr. Omkamal Vashi.

PIERC – 240+ Startups, 1400 Jobs, One National Vision!

Mr. Parth was direct about a limitation that most school AI discussions skip over.

Standard text-generation models like ChatGPT and Gemini are excellent at producing text. They are not designed to produce working software applications from scratch. Students who want to build deployable tools need to understand the difference between text generation and full cloud development, and they need to know which platforms exist for which kind of build. He introduced the participants to Google Labs Opal as the bridge between text-generation interaction and small-application deployment. The key distinctions:

• ChatGPT and Gemini: Strong at text generation, conversational interaction, content drafting. Not designed for application code generation or deployment.
• Google Labs Opal: Environment for building small functional applications using pre-built workflows and automated processes. Ideal for student-level rapid prototyping.
• Traditional cloud architecture (AWS, GCP, Azure): Required for complex applications including video analytics, drone footage processing, or large-scale data systems.
• Claude AI (Anthropic): Useful for complex code generation and structured technical builds, including telematics and dashboard systems.
• Premium tier subscription cost: Generally Rs. 1,500 to Rs. 2,000 per month for the premium tiers of major AI platforms.

To convert the abstract framework into a concrete demonstration, Mr Devariya built a working agricultural disease detection application live.

Opening Google Labs Opal on screen, he typed a detailed instruction: accept a photograph of a crop, identify any disease present, provide a prioritised list of treatment options, and prefer cultural and organic interventions over chemical treatments. Within approximately one minute, Opal produced a fully functional application that he named CropCare. The audience watched as he uploaded a sample photograph of a plant leaf into the application. The system processed the image, matched the visual anomalies against agricultural disease datasets, and generated a detailed report that correctly identified the condition as wheat stripe rust. The treatment list was prioritised exactly as the original prompt had specified, with cultural controls listed first and chemical interventions only as a last-resort recommendation.

The demonstration carried multiple lessons for school participants thinking about real-world applications:

• Application speed: From conceptual specification to a working deployed application in under one minute.
• Domain accuracy: Correct disease identification (wheat stripe rust) from a sample leaf image, demonstrating that current AI platforms have functional capability across agricultural and biological domains.
• Prompt discipline: The output mirrored exactly what the prompt specified, including the priority ordering of cultural over chemical treatments. Prompt quality directly determines output quality.
• Deployment: The application was deployed and made accessible via a shareable URL within the same demonstration window. Students typed the URL into their mobile phones and ran the analytical tool in their browsers. If you wish to master all these tech dimensions and want to work for Big 4s, delay not and enrol in the Tech CSE program of Parul University & say yes to your techie dreams!

Mr. Devariya then demonstrated a more complex build using Claude AI from Anthropic. The target was a vehicular telematics and safety dashboard application that would integrate with a car’s electronic control unit and internal hardware to monitor vehicle status in real time. The demonstration showed a simulated operation: when the car was working normally, the system displayed a vibration pattern consistent with normal road driving. The system was instructed that a sudden spike in energy and vibration would indicate an accident, at which point emergency alerts and GPS location data would be sent automatically to designated emergency responders. He then issued an additional prompt to add computer vision analysis alongside the hardware telemetry. The application’s code was updated within the same demonstration window to incorporate dashboard camera analysis, road condition assessment, and integration with the vehicle’s infotainment system for unified driver controls.

The lesson was structural. Complex multi-system integrations that previously required weeks of engineering work can now be specified, generated, and iterated within a single session.

Mr. Devariya then turned the session toward a topic that most school AI workshops do not address: what happens to the data you upload.

The framing was direct: when a product is free, the user’s personal data is what gets sold. He explained the operational difference between free and premium AI tier data handling:

• Free tier (ChatGPT, Gemini, others): All user inputs (including bank statements, research data, source code, personal documents) are used to train the underlying models. Once data is uploaded, it remains with the service provider as training material.
• Premium tier: Some platforms offer opt-out controls for data ingestion in their premium subscriptions. However, any data uploaded before the opt-out was activated remains with the service provider.
• Premium subscription cost: Typically Rs. 1,500 to Rs. 2,000 per month for the major Western AI platforms.

He then connected the data flow question to national data sovereignty. ChatGPT’s servers are based in the United States. When an Indian person or institution uploads data to ChatGPT, that data travels directly to US-hosted infrastructure. OpenAI has agreements with the United States government that allow American authorities to access user data under specific security and investigation conditions. For Indian researchers, professors, and institutions uploading sensitive research, intellectual property, or government-adjacent data, the implication is significant. He referenced the Government of India’s plan to build sovereign data centres within India as the structural response to the foreign-data-residency challenge, aligning with the broader Make in India and Atmanirbhar Bharat priorities on digital infrastructure self-reliance.

Use AI as a supporting agent. AI is not everything.

Ms. Anbumathi M of the Wadhwani Foundation, in her valedictory address, echoed the same caution

Mr Devariya touched on several additional applications, each grounded in a real-world use case students could replicate.

• Deepfake detection app: An application concept for verifying whether photos and videos circulating on social media were genuine or AI-generated, with user-reporting functionality for content removal requests.
• Automated answer sheet evaluation: A system connecting to cloud storage (Google Drive), reading handwritten student answer sheets, evaluating content against curriculum standards, and generating grades with reasoning explanations. Addressed a teacher concern raised during the session.
• Handwriting improvement tool: An application for school-age children that analyses how they form letters and provides specific corrective feedback to improve handwriting quality over time.
• QR-code-based campus information system: A student concept (highlighted during the session) for placing QR codes throughout school facilities; scanning them would deliver contextual content (videos, animations, summaries) about that specific location.
• Algorithmic trading systems: Devariya referenced a case study from a finance and technology conference in which an engineering student deployed Rs. 10 lakh into an automated trading system that generated approximately Rs. 1 crore in one month. The case was offered with appropriate caution about risk and verification.

Mr. Devariya closed with a reference to Nvidia’s Chief Executive Officer Jensen Huang‘s framing of Open-Source Cloud AI at a recent technology conference. The framework describes how companies are now operating digital agents from messaging apps like Telegram and WhatsApp, with automated workflows running continuously in the background to process emails, generate summaries, manage databases, and execute software tasks without direct human intervention. The implication for school students and teachers in the audience was direct: software systems worth billions of dollars can now be built and deployed by people who do not know how to code. The technology democratisation that the next decade will produce aligns directly with the Government of India’s Startup India priority of expanding entrepreneurial access beyond traditional technology backgrounds.

The AI session was strategically positioned on Day 2 of the programme. The morning session by Mr. Mithilesh Patel of Vraj Innovator covered the structural framework for converting ideas into ventures: problem identification, solution design, material feasibility, prototype development, testing, market strategy, team building, intellectual property, and budget planning. Mr Devariya’s afternoon session then provided the technology toolkit through which students could execute that framework. The Day 1 design thinking foundations laid by Dr Arvind Deshmukh and Ms Anbumathi M provided the methodology that drove the technology demonstrations on Day 2, and the Day 3 final pitch session (where students presented 50 innovation projects) was the evaluative culmination.