AI / ML Engineer
Speech systems, grounded LLMs, and backends that stay within the hardware.
I study computers and artificial intelligence at Fayoum University and build systems that stay honest about their evidence: speech pipelines for Egyptian Arabic, retrieval-augmented generation, and APIs that fit the GPUs they run on.
Training path: machine learning and deep learning, then NLP, then IBM's generative AI specialization. The applied work — GitHub and a live RAG space — is what I put in front of people.
2022–2026
Fayoum University, Faculty of Computers and Artificial Intelligence.
Foundations
Stanford / DeepLearning.AI tracks — from linear models through CNNs and sequence models.
Language
Classification, probabilistic models, sequence models, attention.
Applied
RAG, LangChain, adapters, LoRA, and alignment-style fine-tuning labs.
Tools that appear in the repositories — not a wishlist.
Production systems, speech analytics, and research implementations.
Speech analytics, neural pipelines, and applied machine learning models.
Chunk a PDF, embed locally with ONNX MiniLM, retrieve in Chroma, and answer only from the document via Groq.
Egyptian Arabic session pipeline: emotion, ASR, sentiment, and report generation behind a FastAPI service.
sklearn Pipeline and GridSearchCV over classical models on DIDA digits, with reproducible preprocessing.
Fine-tuning track through LoRA, QLoRA, instruction tuning, reward models, PPO, and DPO, plus RAG with LangChain.
From vector spaces and n-grams through sequence models, NER, attention, translation, and question answering.
No invented client quotes. These are the constraints I actually use.
If the document does not contain it, the system should say so.
Chunk audio and models so long sessions fit the GPUs you actually have.
Demos ship next to the repository, not as slides alone.
Internships and junior AI, speech, or backend roles. The form opens your mail client with the note filled in — nothing is stored on a server.
Book a callahmed022magdy@gmail.com
+20 102 517 3032
Find my open-source code, research milestones, and technical posts.