Re: GSoC 2026: Prototype & Contribution Inquiry – Gemini Care Assistant

[Tanisha Ray (B24CM1061)]

Hello Catrobat team,

My name is Tanisha Ray, and I am a B.Tech student in Artificial Intelligence and Data Science at IIT Jodhpur, India. I am very interested in the GSoC 2026 project idea "Gemini-API-Powered Intelligent Care Assistant." After reading the project description and the ongoing discussions here, I find the focus on responsible AI systems and human-in-the-loop decision support particularly compelling.

From my understanding, the project aims to build an AI interpretation layer that converts structured IoT activity signals (movement logs, time-based routines, and activity confirmations) into meaningful insights for caregivers. The Gemini API would be used to generate contextual explanations while enforcing safeguards such as non-diagnostic prompts, explainability-first responses, role-aware output filtering, and supervisor validation before escalation.

My experience includes building real-time AI pipelines and LLM-integrated backend systems, which I believe align well with the goals of this project:

• Inter IIT Tech Meet – Gold Medal (Real-Time Fraud Detection System): Contributed to a production-grade streaming ML system with real-time inference pipelines and human-in-the-loop decision support, including explainability interfaces for analysts.

• Jarvis - Real-Time Multimodal Conversational AI (College project): Contributed to a real-time voice assistant pipeline (STT -> LLM -> TTS) using FastAPI, WebSockets, and vLLM, focusing on scalable LLM inference and structured AI pipelines supporting 500+ concurrent users.

To explore the technical direction further, I have started sketching a small prototype that simulates:

• ingestion of structured IoT activity logs
• time-windowed routine summarization
• deviation detection from baseline activity patterns
• Gemini-based explanation generation with strict non-diagnostic prompts
• role-based output filtering with supervisor validation workflows

Since there isn't a dedicated starter task for this idea, I wanted to ask if there are recommended areas where students can start contributing before submitting their proposals.

In particular, I would appreciate guidance on:

1. Whether there are existing repositories, datasets, or prototypes related to this idea.

2. Whether applicants are encouraged to build small proof-of-concept prototypes before proposal submission.

I would be happy to start exploring the ecosystem and contributing wherever helpful while preparing my proposal.

Thank you for your time and guidance.

Best regards,
Tanisha Ray
IIT Jodhpur, India