GSoC 2026 Pre-Coding: Preparation & Onboarding

Read two papers shared by Hizkeal on DBpedia (DBpedia Paper 1, DBpedia Paper 2).

Read the three papers on XLM-R, Afro-XLM-R, and BERT.

Started refactoring the Amharic Dbpedia Website locally with vanilla Typescript (repo).

Participated in the DBpedia GSoC 2026 meeting on May 22, 2026. Although the session was shorter than anticipated, the mentors were highly collaborative and provided clear, direct insights.

Gained a comprehensive understanding of the other contributors' project scopes.

Documented the progress achieved throughout the week in a formal report.

I had my first onboarding meeting on Thursday with my GSOC Mentors (Dr. Richard, Hizkeal, Tilahun, and Andargachew). We introduced ourselves to each other and decided to have our meeting be on Fridays at 2:00 PM.

They also gave me to checkout the current Models that being used for Ontology Alignment and property retrieval on the Amharic DBpedia.

Joined and integrated into all primary communication channels: Slack, WhatsApp, and Microsoft Teams.

Conducted a revision of the Amharic DBpedia Chapter research paper (Link).

I have read and performed model benchmarking on Kaggle using GPU access to explore mBERT, XLM-R, and Afro-XLM-R that was developed by the DICE-Research team (Kaggle).

Revised core knowledge of Transformers, BERT, and efficient fine-tuning methods like LoRA (Notes).

Selected LangGraph as the primary framework for building the project's agentic orchestration layer and read about Langgraph (Docs).

Why LangGraph? It is used by production-grade companies like LinkedIn. It has large community support. It is customized for both python and javascript compared to Mastra.

Developed and launched a personal portfolio website and started writing a blog about GSOC progress each week.

Prepared a technical reference document for mentors to outline the proposed development roadmap.