A Google Developer Group (GDG) on Campus — JIS University session
Build real AI applications, not just prompts.
Beyond the Chatbot was a GDG on Campus JIS University learning session designed to take students beyond experimenting with prompts and introduce the process of building practical, reliable AI-powered applications.
The session explored how Large Language Models (LLMs) can be integrated into applications, how to work with the Gemini API, and what it takes to move from a working prototype to a more dependable application.
The event also featured an Introduction to GDG on Campus at JIS University, giving participants an opportunity to learn about the developer community, connect with fellow students, and understand ways to get involved.
| Detail | Information |
|---|---|
| Event | Beyond the Chatbot |
| Theme | Building and Shipping Real AI Apps with LLMs |
| Community | Google Developer Group on Campus — JIS University |
| Date | 5 November 2026 |
| Time | 2:00 PM – 5:00 PM |
| Venue | JIS University, Kolkata |
| Speaker | Suryanshu Paul, AI Engineer at humaineeti |
The session focuses on the practical building blocks of modern LLM applications:
- Understanding Large Language Models and how they are used in applications.
- Tokens, context windows, temperature, and hallucinations.
- Understanding the strengths and limitations of model-generated responses.
- Structuring prompts with roles, context, rules, examples, and output formats.
- Improving consistency and clarity in model responses.
- Knowing when prompt engineering is enough and when a more advanced approach is needed.
- Connecting applications to an LLM through an API.
- Working with the Gemini API and Python.
- Handling model responses in application code.
- Using structured output to make responses easier to validate and process.
- Structured JSON output.
- Retrieval-Augmented Generation (RAG).
- Embeddings and retrieval.
- Function calling and application-controlled tools.
- An introduction to AI agent workflows.
- Building assistants around useful information sources.
- Summarising and extracting information from documents.
- Designing workflows that are useful beyond a single prompt.
- Thinking about reliability, security, latency, and cost.
Want to revisit the concepts discussed during the session?
The complete presentation is available below.
Beyond the Chatbot – Session Presentation
Feel free to download the slides and use them as a learning resource.
Speaker: Suryanshu Paul
Community: GDG on Campus – JIS University
A practical example discussed in the session is a Campus Notice Assistant built with Python and the Gemini API.
The idea is to help students work with campus notices by turning unstructured notice content into concise summaries and structured information.
The walkthrough is organised around these steps:
- Summarise a notice — generate a short summary in three bullet points.
- Extract event information — identify fields such as event title, date, and venue in a structured JSON response.
- Use function calling — demonstrate how a model can request an application function, such as adding an event to a calendar.
- Improve reliability — consider retries with exponential backoff and a fallback model.
- Add observability — track details such as the model used, response latency, and token usage.
The session material explains the demo workflow and the engineering concepts behind it. Refer to the source code present in this repository for the exact implementation and available features.
A successful model response is only one part of a real application. Production-minded development also means considering:
- Validation: Check model output before your application uses it.
- Reliability: Handle timeouts, transient errors, retries, and provider failures.
- Fallbacks: Have a sensible alternative when a model or service is unavailable.
- Evaluation: Test responses against realistic examples rather than relying on a few successful prompts.
- Observability: Measure latency, token usage, errors, and cost.
- Security: Protect API keys and treat user-provided documents as untrusted input.
- Human oversight: Require approval before performing sensitive or irreversible actions.
- Start with a simple prompt and add complexity only when the use case requires it.
- Use RAG when the model needs to answer using external or domain-specific information.
- Keep tool execution under application control.
- Validate structured output before using it.
- Measure quality, latency, and cost as part of development.
- Design for failure instead of assuming every API call will succeed.
GDG on Campus is a student developer community that creates opportunities to learn, build, collaborate, and connect with the wider developer ecosystem.
The JIS University community aims to bring students together through technical sessions, hands-on learning, peer collaboration, and community activities.
The event also included an introduction to GDG on Campus JIS University and its community.
This marks an important step for the community as the new organizing team takes forward its activities, learning initiatives, and student engagement.
We congratulate the new organizers and look forward to the sessions, workshops, collaborations, and opportunities they will help create for the JIS University developer community.
The photograph is shared as an event record. Individual names and roles are not listed here.
Whether you are just starting with AI or already building applications, use this as a practical path after the session.
| Stage | Focus | Try this |
|---|---|---|
| 1 | Python fundamentals | Practise functions, loops, and file handling. |
| 2 | Gemini API | Make a first model call from a Python script. |
| 3 | Prompting | Build a notice or lecture-notes summariser. |
| 4 | Structured output | Extract dates, venues, and titles into JSON. |
| Stage | Focus | Try this |
|---|---|---|
| 1 | RAG | Build question answering over syllabus PDFs. |
| 2 | Function calling | Connect a model to a calculator or calendar function. |
| 3 | Reliability | Add timeouts, retries, and a fallback. |
| 4 | Evaluation | Test the app with a small set of realistic questions. |
| 5 | Deployment | Deploy the project and monitor usage and errors. |
The examples and concepts are based on Python and the Gemini API.
- Python 3.10 or later.
- A Google account.
- A Gemini API key from Google AI Studio.
- Basic familiarity with Python and the command line.
git clone <YOUR_GITHUB_REPOSITORY_URL>
cd <YOUR_REPOSITORY_FOLDER>Replace the placeholders with your repository URL and folder name.
Windows (PowerShell):
python -m venv .venv
.venv\Scripts\Activate.ps1macOS / Linux:
python3 -m venv .venv
source .venv/bin/activateThe session uses the Google Gen AI Python SDK and Pydantic for structured output.
pip install -U google-genai pydanticIf a requirements.txt is included in the repository, you can install its dependencies with:
pip install -r requirements.txtSet your API key as an environment variable. Never hard-code it in source files or commit it to Git.
Windows (PowerShell):
$env:GEMINI_API_KEY="YOUR_API_KEY"macOS / Linux:
export GEMINI_API_KEY="YOUR_API_KEY"Get your API key from Google AI Studio.
Use the entry-point filename and instructions provided by the actual source code in this repository. For a Python entry file, the command will generally look like:
python <entry_file>.pyContributions that improve the examples, documentation, code quality, tests, or learning experience are welcome.
- Fork this repository.
- Create a feature branch.
- Make your changes and test them.
- Open a pull request with a clear description.
Please do not commit API keys, credentials, private notices, or other sensitive information.
A big thank you to Suryanshu Paul for sharing practical insights into building and shipping real AI applications, and to everyone who contributed to making the GDG on Campus JIS University session possible.
Thank you to the organizers, volunteers, and participants who made the event a space for learning, collaboration, and community.
Google Developer Group on Campus — JIS University
Learn. Build. Connect.
Beyond the Chatbot — Build real AI applications, not just prompts.