I'm an AI Engineer with a B.Sc. in Artificial Intelligence & Data Science (graduation project: A+ / 99%) and an AWS Certified AI Practitioner. I take LLM, RAG and agentic systems from experiment to production: retrieval design, fine-tuning & quantization, evaluation gates, observability and deployment.
I validate with real metrics, not vibes: RAGAS gates in CI, load tests with P95 latency, drift detection, leakage audits, and benchmarked quantization.
| 🏗️ Ship | Full products end to end: LLM pipelines, FastAPI backends, Dockerized deployments, live demos |
| 🔬 Measure | MLflow experiments, RAGAS, perplexity, tokens/s, VRAM, Locust load tests |
| 🌍 Arabic NLP | Arabic legal RAG, Arabic poetry LLM, Arabic OCR fine-tuning |
| 🎓 Teach | AI & LLMs instructor for online learners |
- 🔤 Fine-tuning a small, efficient Arabic OCR model (TrOCR, LoRA/PEFT)
- 🧪 Production-grade RAG pipelines with real retrieval metrics
- 🤖 Multi-agent workflows and tool-use patterns
- 🏗️ LLMOps: MLflow, ONNX, monitoring, CI quality gates
- 📚 Sharpening DSA and system design
🏗️ CadArenaNatural language (Arabic/English) → CAD (DXF). Turns plain-language engineering commands into valid floor plans, with an Egyptian Building Code (EBC 2023) quality gate and a RAG module (5 parsers: PDF, DXF, IFC, CSV, XLSX, knowledge graph, validation agents). 85–90% less design time vs. manual drafting. 255 commits, 40+ API endpoints, 120+ tests, live demo. 🎓 A+ (99%).
|
⚖️ Sanad (سند)Arabic legal RAG over all 1,149 Egyptian Civil Code articles. Hybrid retrieval (Qdrant dense + BM25 + exact-article lookup, RRF k=60), grounding guardrail that refuses without a supporting article. 8 configs compared in MLflow, RAGAS CI gate (faithfulness 0.90 vs 0.75), P95 418 ms @ 49.4 RPS, 0% errors, drift detection, 5% canary routing, PII scrubber.
|
🥗 NutriAIFull-stack RAG SaaS (114 commits). PDF/TXT ingestion → semantic chunking → embeddings → vector search → grounded LLM answers. Switchable providers (OpenAI/Cohere), JWT auth, per-user knowledge bases, Prometheus metrics.
|
Paste any public GitHub repo → architecture summary, interactive dependency graph (files → classes → functions), hybrid static + LLM bug detection, refactoring diffs and RAG Q&A with citations. Clean Architecture, Celery workers, WebSocket progress.
|
|
Arabic poetry LLM. QLoRA fine-tune (NF4, r=16) of Qwen2.5-1.5B-Instruct on an Ashaar-based instruction set (10 themes) using a single T4. Published on Hugging Face with a streaming Gradio demo Space.
|
Open reference notebook, 16 sections: from-scratch affine/group-wise quantization, GPTQ, AWQ, GGUF K-quants, NF4/LLM.int8, HQQ, SmoothQuant, KV-cache, QLoRA, QAT. Benchmarked on size, Wikitext-2 perplexity, tokens/s and peak VRAM.
|
|
Config-driven pipeline on 568K transactions: leakage audit (dropped an ID feature with r≈0.86), stratified 5-fold CV, metric-based quality gates (PR-AUC, min precision/recall) across LR / RF / XGBoost. MLflow registry, FastAPI + Pytest, Docker, ONNX export + inference benchmark.
|
⌨️ yctl-cliPersonal AI engineer CLI for Ubuntu: scaffolds NLP/CV/ML projects, inspects datasets (CSV, Excel, JSON, Parquet) and diagnoses Python/CUDA/GPU health.
|
| Project | What it does | Stack |
|---|---|---|
| 💊 PharmaSaaS | Multi-tenant pharmacy platform: inventory, orders, prescriptions with branch-level data isolation | FastAPI React 19 PostgreSQL |
| 🔧 CadArena DXF Viewer | Companion viewer for CadArena DXF outputs | Web |
| 🛰️ Aether Analytics | NASA Space Apps 2025 (2-person team): automated retrieval, cleaning and visualization of NASA earth/space datasets | Python Data Viz |
| 🎙️ Speech Emotion Recognition | CNN classifier for emotion in audio using MFCC/spectrogram features | CNN Kaggle |
| 🫁 Pneumonia X-ray Classifier | MobileNetV2 transfer learning + Gradio app | TensorFlow Gradio |
| ⚡ Big Data with PySpark | Customer segmentation + Kaggle TPS Apr-2021 baseline | PySpark |
| 🔍 Vector DB Search | Semantic and image search with Pinecone and Qdrant | Pinecone Qdrant |
| 🧭 AI Search Algorithms | Hill Climbing, Simulated Annealing, UCS, IDFS, maze game with AI search | Python |
| 📊 Dash Dashboard · 🖼️ Image Processing GUI | Data-visualization dashboard and image-processing desktop GUI | Dash Python |
🔗 All repos: github.com/Youssefx64?tab=repositories
| Languages |     |
| GenAI / LLM |      |
| ML / DL |     |
| Backend / Infra |     |
| Vector DBs |     |
| NLP |    |
| Frontend |    |
| MLOps |      |
| Tools |     |
Also: Hugging Face (Transformers, PEFT, TRL) · RAGAS · Langfuse · Locust · BentoML · BM25 / RRF / reranking · GPTQ / AWQ / GGUF · vLLM · TorchServe · Celery · Redis · WebSockets · JWT
 |
 |
 |
 |
B.Sc. in Artificial Intelligence & Data Science, Faculty of Computers & AI, Beni-Suef National University (Oct 2022 – Jul 2026)
- 🏆 Graduation Project: CadArena, score A+ (99%)
- 📖 Machine Learning · Deep Learning · NLP · Computer Vision · Big Data · Data Structures & Algorithms · AI Search
| Role | Organization | Period |
|---|---|---|
| Backend AI Engineer Intern | FlyRank AI (Remote) | Jun 2026 – Oct 2026 |
| Founder & Team Lead | CadArena (Capstone) | Sep 2025 – Jun 2026 |
| AI/ML Intern (ML, CV & NLP tracks) | National Telecommunication Institute (NTI) | Dec 2024 – Sep 2025 |
| AI & LLMs Instructor | Self-Employed, Remote | Oct 2025 – Nov 2025 |
| Big Data Engineer Intern | Huawei | Aug 2025 |
| Data Engineer Trainee | Digital Egypt Pioneers Initiative (DEPI) | Jun 2024 – Sep 2024 |
| AI Fundamentals Trainee | Zewail City of Science and Technology | Aug 2023 – May 2024 |
AWS Certified AI Practitioner · NVIDIA · Building LLM Applications With Prompt Engineering · NVIDIA DLI · Generative AI · NVIDIA · AI for All · Stanford & DeepLearning.AI · Supervised ML · Oracle OCI 2025 AI Foundations Associate · NTI · NLP · NTI · Deep Learning · Microsoft Data Engineer (DEPI) · Huawei Big Data Associate · Google · Build with AI · NASA Space Apps Challenge 2025