Open source Contribution in Gsoc 2026

[Ashitpatel001]

Hi OpenVINO Team!

I’m Ashit Patel, a Computer Science undergrad with experience in Python, Deep Learning, and RAG pipelines (LangChain/Gemini). I’ve gone through the Contribution Guide and am eager to start my prerequisite task for GSoC 2026.

I am particularly interested in the GenAI / Python API segment. Could you point me toward any unassigned 'Good First Issues' or a specific module that needs testing/fixing right now?

Looking forward to contributing! Github link

[masonearl]

Welcome Ashit! Great to see you interested in OpenVINO GenAI.

Getting Started with GenAI / Python API

1. GenAI repo - Check openvinotoolkit/openvino.genai for GenAI-specific work.

2. Good first issues - Filter by good first issue in both the main repo and genai repo.

3. Run the GenAI samples - Start with the GenAI API samples to understand the API.

Unassigned Issues

Check these filters:

• Good first issues (unassigned)
• Python API issues

Tool Recommendation

For exploring large codebases, try Cursor IDE - it's VS Code with AI that can answer questions about the code. You can ask "how does the GenAI API handle tokenization?" and it'll find and explain it.

Good luck with GSoC 2026!

[Ashitpatel001]

Hi @masonearl

From last few weeks i have explored openvino.genai and i wanted to share a quick update on my progress.

I have successfully built and run the Live VLM Chat (C++) sample locally. Getting the C++ environment, OpenCV, and model weights set up gave me a strong understanding of the low-level pipeline specifically how openvino.genai handles tokenization and inference loops on the backend.

Now that I’ve verified the "engine" works, I am shifting focus to Project #1 (GUI Agent).

My Plan: I believe the "Agent" and "GUI" layers are best implemented in Python for flexibility, while relying on OpenVINO's C++ bindings for performance. I am starting work on a Python-based prototype (likely using Streamlit or PyQt) that wraps this VLM pipeline into a user-friendly interface with basic history/context.

Does this hybrid approach (Python GUI + OpenVINO Backend) align with what you envision for the GSoC project?

Thanks!