Building AI Applications With Gemini: What the Goa Cloud Connect 2026 Baseline Session Taught 120 Developers at Parul University, Goa

Most people meet generative AI as a chatbot. Developers meet it as a model, an API, and a deployment target, and the distance between those two experiences is exactly what a good workshop closes. On 11 July 2026, the Faculty of IT and Computer Science at Parul University Goa hosted the opening session of Goa Cloud Connect 2026, a four-week hands-on series organised by Google Developer Group (GDG) Goa under the theme "From Zero to Production in 4 Weeks". The first session, Hello Gemini, Setting the Baseline was built around a single practical question: what does building AI applications with Gemini actually involve once you move past the demo?

For the 120 students, developers, and technology enthusiasts in the seminar hall, the answer was not a lecture on artificial intelligence in the abstract. It was a working baseline.

The baseline is not a demo. It is the point where a prompt becomes a product.

The session covered the foundations of generative AI using Google's Gemini models, Vertex AI, multimodal capabilities, and prompt engineering. Each of those is a specific, learnable thing, and understanding what each does is the difference between a student who has heard of AI and one who can build with it.

• Gemini is Google's family of large multimodal models, the reasoning engine that generates text, analyses images, and works across formats.
• Vertex AI is the Google Cloud platform that lets developers access those models, tune them, and put them into applications through a managed interface rather than raw infrastructure.
• Multimodal capability means a single model can take text, images, audio, or a combination as input, which is what separates current systems from the text-only tools that came before.
• Prompt engineering is the discipline of instructing the model precisely enough to get reliable, useful output, and it is the first real skill a builder needs.

Set out this way, the "baseline" is a stack, not a slogan. A developer who understands how Gemini, Vertex AI, and prompt design fit together has the vocabulary and the mental model to build the rest. That is what a foundational session is for, and it is why the series began here rather than with a finished product.

The technical content was anchored by Ashutosh S. Bhakare, a Google Developer Expert for Google Cloud and a Docker Captain, who is CEO of Unnati Development and Training Centre and a DevOps architect with over two decades of experience. Both of those recognitions are selective. Google Developer Experts are independently recognised by Google for demonstrated expertise, and Docker Captains are a small group acknowledged by Docker for technical depth and community contribution.

The reason this matters is the credibility of instruction. A foundational AI session led by someone who has spent years shipping and deploying real systems is qualitatively different from one delivered from slides alone. Bhakare's background in DevOps and containerisation also previews where the series was heading, because the later sessions move from building a model-powered application to grounding it in real data and finally deploying it to production on Cloud Run and Firebase.

The session was coordinated on the faculty side by Dr. Shailesh Gahane, with the wider program organised by GDG Goa lead Vedanth Bandodkar alongside Priyal Nagvekar of Parul University Goa, supported by student volunteers Shriyansh Narvekar, Yash Mukade, and Brahmay Surlakar.

There is a predictable failure pattern in how students approach AI. They jump to the impressive end, autonomous agents and full applications, without the baseline, and then cannot debug what they build because they never understood the parts. The order of this series is a correction to that pattern.

Session one establishes what a model is and how to instruct it. The second session grounds that model in real-world data using enterprise RAG and vector search. The third moves to autonomous, multi-agent workflows. The fourth deploys a full AI-native product to production. Each session assumes the one before it, which is exactly how professional capability is built, and exactly what is missing from most self-taught AI learning.

For a student comparing where to study computer science, this sequencing is the point worth noticing. A curriculum, or a campus series, that teaches AI in the right order produces builders. One that teaches only the exciting parts produces people who can describe AI but not ship it.

The 120 attendees did not leave with only concepts. The program provided Google Certificates, Google Cloud Credits, one-on-one mentorship, and direct networking with working professionals. Those are not incidental perks. Cloud Credits let a student actually build on Vertex AI rather than only read about it, and one-on-one mentorship with a Google Developer Expert is the kind of access that ordinarily costs money or connections.

This is also where the workshop connects to a wider pattern on the Goa campus. The Faculty of IT and Computer Science has been putting students in front of real technology and real institutions rather than keeping them in the classroom, from an industrial visit to D-Link's networking plant in Goa to a research visit inside NCPOR, India's polar research nodal agency, where students saw data science applied to real climate research. The Cloud Connect series sits in that same approach: exposure to the real thing, taught by people who do it.

The session was built against defined objectives: to give participants an understanding of cloud computing, AI-native application design, and modern deployment on platforms such as Vertex AI and Cloud Run; to familiarise them with industry-standard development workflows; to explain the practical applications of Gemini models, multimodal systems, enterprise RAG, and autonomous agent workflows; to expose them to emerging research and innovation in AI and digital infrastructure; and to build awareness of careers in cloud architecture, DevOps, and artificial intelligence.

The program mapped to four United Nations Sustainable Development Goals: SDG 4 on quality education through hands-on learning, SDG 8 on decent work and economic growth through exposure to the cloud and AI job market, SDG 9 on industry, innovation and infrastructure through modern cloud architecture, and SDG 12 on responsible consumption and production through efficient, scalable resource management on cloud platforms.