class MagdiWaleed:
def __init__(self):
self.role = "AI / ML Engineer"
self.company = "Nabrah.ai"
self.location = "Riyadh, Saudi Arabia"
self.focus = ["Speech Recognition (ASR)", "Agentic AI", "LLMs", "MLOps"]
self.lifecycle = "research → fine-tuning → deployment → scale"
self.languages = ["Arabic (native)", "English (professional)"]
def current_mission(self):
return "Shipping real-world AI with measurable impact 🚀"- 🎙️ I build production speech-recognition systems for Arabic & English, optimized for real-time, low-latency inference.
- 🤖 I design agentic systems multi-step reasoning, tool use, and workflow automation with LangChain / LangGraph.
- ☁️ I work across the full ML lifecycle: from model training and fine-tuning to cloud-native backends on GCP.
- 🎓 B.Sc. in Computer Science (AI) Cairo University, Faculty of Computers & AI.
- 🎬 I run AI Student Journey on YouTube teaching coding & AI from zero.
- 💬 Ask me about ASR, Triton inference, LLM agents, or deploying ML at scale.
|
Production-grade Arabic speech-recognition system with bilingual Arabic/English (code-switching) support. Streaming, low-latency inference served via Triton on GCP with GPU acceleration. Highlights: outperforms Whisper large-v3 by 41%+ WER and 19× RTF.
|
Real-time Turkish Sign Language translation. A multi-fusion model trained from scratch for gloss-level translation from mobile camera input, wrapped with boundary detection + NLP for continuous translation.
|
|
Agentic customer-service assistant with context retrieval, multi-step reasoning, and workflow automation. Real-time multi-user support, conversation memory, and automated service recommendations.
|
AI-powered market analyzer for the Saudi market benchmarks products, identifies market gaps, and generates actionable recommendations through an interactive interface.
|
- DeepLearning.AI TensorFlow Developer Coursera (2024)
"Driven by building real-world solutions with measurable impact."