Skip to content
View Ruqyai's full-sized avatar
💻
💻

Block or report Ruqyai

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Ruqyai/README.md

Ruqiya Bin Safi · رقيا بن صافي

AI engineer and researcher. I build production LLM systems, study what happens inside them, and make them work in Arabic.

Now: AI safety research with EleutherAI on hierarchy in sparse autoencoders, continuing after the SOAR 2026 program (July–August 2026).

Before: AI Safety Research Fellow at Algoverse (2026), and nearly three years at Fyler building a GenAI search engine for MENA, from the first architecture to product ownership.

Website · عربي · Certificates · LinkedIn · Hugging Face · Kaggle · X

Selected work

Project What it is
MMTEB · ICLR 2025 Massive Multilingual Text Embedding Benchmark. Co-author; contributed Arabic evaluation data (mteb#710).
ChatBard One of the first Arabic apps on PaLM 2. Featured by Google Cloud Tech.
Ruqia-Library Arabic text cleaning, normalization and diacritics. pip install ruqia
RuqLM Small Arabic language models, 3M to 30M parameters. Weights
Gemma 2 for Arabic Gemma 2 2B and 9B fine-tuned on Arabic data with LoRA.
Marketing agents Agents that write and post LinkedIn content, built with Google ADK and A2A on Gemini 3 Pro.

Talks and courses

130+ talks, workshops and mentoring sessions in 10 countries since 2017. See the talks.

Courses in Arabic:

Community

2× Google Developer Expert (AI, Google Cloud) · Google Cloud Champion Innovator · TensorFlow Contributor Award 2021, Middle East & Africa

Community builder: founded, led and grew tech communities for developers and for women in tech in Saudi Arabia for five years (2019 – 2024), as an organizer, trainer and mentor.

I write about LLMs, Arabic NLP and AI products on my blog.

Pinned Loading

  1. Deep-Learning-Course Deep-Learning-Course Public

    Deep learning course

    Jupyter Notebook 19 12

  2. Amaken Amaken Public

    Senior Project

    Java 2 1

  3. ChatBard ChatBard Public

    ChatBard: An Intelligent Customer Service Center App Using PaLM2 APIs

    Jupyter Notebook 49 9

  4. Data-Engineering-Course Data-Engineering-Course Public

    A data engineering course that focuses on how to create a real-time Twitter analytics dashboard for streaming tweets with a sentiment analysis (NLP) feature.

    Jupyter Notebook 3 13

  5. Machine-Learning-Projects-For-Production Machine-Learning-Projects-For-Production Public

    HTML 2 1

  6. Ruqia-Library Ruqia-Library Public

    Python library used for Arabic NLP to process, prepare and clean the Arabic text

    Python 16 6