Skip to content
View khir00's full-sized avatar
🎯
Focusing
🎯
Focusing

Highlights

  • Pro

Block or report khir00

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

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

Report abuse
khir00/README.md

Hi there 👋

Mohamad Khir Alhomsi (Khir)

AI Engineer — GenAI & LLM Applications

I build production GenAI systems (RAG, agents, fine-tuning) — and document the path from AI grad to AI Engineer, in Arabic and English.

🎓 B.Eng. Information Engineering (AI & Data Science), Syrian Private University 💼 AI Engineer @ AppTech Systems

🔧 Featured Work

AI Career RAG — Bilingual (Arabic/English) RAG career assistant Grounded retrieval + generation, evaluated systematically — not a chatbot wrapper. Precision@1: 80% · Precision@3: 100% · SBERT + FAISS + Qwen2.5-1.5B 🔗 Live Demo

SignStream — ASL-to-speech pipeline (video → sign → English → audio) Efficient pose-based Transformer, fine-tuned T5 for gloss-to-text. 74.70% val / 67.40% test top-1 · 18.65M params · 2,731 classes · ASL Citizen (~83K videos)

Diabetic Retinopathy Grading — Ordinal severity classification EfficientNetB3 + CORAL ordinal regression, deployed as clean inference package. QWK 0.9143 · 81.86% accuracy · weighted F1 0.824 · APTOS 2019

📍 Currently

Building GenAI systems and writing the AI-grad → AI-Engineer roadmap for Arabic-speaking AI graduates.

📫 Connect

LinkedIn

Pinned Loading

  1. Diabetic-Retinopathy Diabetic-Retinopathy Public

    Diabetic retinopathy grading (0–4) from retinal fundus photos — ordinal CORAL + EfficientNetB3, 0.9143 QWK on APTOS 2019.

    Jupyter Notebook

  2. SignStream- SignStream- Public

    Efficient pose-based Transformer for real-time ASL recognition — video → sign → English → speech. 18.65M params, 67.40% test top-1 on 2,731-class ASL Citizen.

    Python

  3. ai-career-rag ai-career-rag Public

    Bilingual (Arabic/English) RAG system for AI career guidance — grounded retrieval + generation, evaluated across baseline, upgraded embeddings, and re-ranking stages

    Python