Al-Hossein Mahmoud
AI & ML Engineer specializing in building intelligent systems
including Ai & ML, NLP projects, and Agentic Engineering to clean web interfaces, Passionate about turning complex ideas
into practical, working
solutions.
Who I Am
I'm an AI & ML Engineer focused on building intelligent systems that solve real-world problems, not just technical demos. I believe the best AI architecture isn't about using the most complex model, but choosing the right tool for the job.
My expertise spans Multi-Agent Systems, RAG Architectures, and Privacy-First ML. I recently engineered a fully local, 8-agent clinical decision support pipeline that prevents AI hallucinations without relying on cloud APIs. Beyond code, I've shipped production-ready web apps, reached the finals of the EVA and IEEE ICIAAI AI Hackathons, and founded my own clothing brand — I thrive at the intersection of deep tech and creative execution.
I am currently seeking AI/ML Engineering roles, internships, and collaborations where I can build impactful, scalable, and transparent AI systems.
Academic Background
Tech Stack
What I've Built
An advanced 8-stage clinical AI agent pipeline designed to augment healthcare workflows and clinical decision support. Features automated intake normalization, NLP-based medical entity extraction, ML-driven risk stratification, RAG-powered evidence retrieval, drug-interaction checking, clinical guideline verification, multi-step clinical reasoning, and automated final report generation. Built for high accuracy, safety, and explainability.
Team capstone for the NTI NLP track. An end-to-end app that classifies customer complaints by intent and sentiment, clusters them by topic, and drafts a suggested reply. Trained TF-IDF + Logistic Regression for intent classification and MiniLM embeddings + KMeans for topic clustering; integrated pretrained sentiment and reply-generation models.
An end-to-end LLMOps pipeline built on Google Cloud to fine-tune and deploy a specialized ML/AI engineering Q&A assistant. Ingests curated Stack Overflow data via BigQuery, orchestrates reproducible instruction-tuning of a foundation model using Vertex AI and Kubeflow Pipelines, and serves predictions via a FastAPI endpoint featuring strict prompt-template consistency and dual-layer safety/citation gates.
A 7-class dermatoscopic image classifier trained on the HAM10000 dataset, designed as a screening-aid prototype. Features a robust ML pipeline with lesion-grouped stratified splitting to prevent data leakage, class-weighted loss, and optimization for high-risk-class recall. Includes Grad-CAM interpretability overlays and a FastAPI backend with a lightweight frontend UI for real-time image upload and prediction.
An end-to-end fake news classification system using a fine-tuned DistilBERT model trained on the WELFake dataset. Features a FastAPI backend for real-time inference (returning verdicts with confidence probabilities) and a lightweight frontend interface ("Wire Desk") for users to submit article titles and text for instant Fake vs. Real verification.
Developed during a lab internship at Hindawi, this Retrieval-Augmented Generation (RAG) chatbot enables intelligent, context-aware querying over specialized document collections. Engineered an end-to-end pipeline for document ingestion, text chunking, and vector embedding, coupled with an LLM to generate accurate, grounded responses. Designed to streamline information retrieval and enhance user interaction with complex,domain-specific textual data.
Where I've Worked
Intensive NLP track covering the field end-to-end: text preprocessing, word embeddings, sequence models, Transformers, LLMs, and fine-tuning — applied through hands-on classification and sentiment analysis pipelines. Capped off with a solo capstone: the Customer Complaint Analyzer (see Projects).
Working on applied LLM and agentic systems. Shipped a RAG-based agent, YouTube video summarization pipeline and working on LangChain, contributing end-to-end from pipeline design through deployment.
Advanced to the Finals as part of a team, ranking among the top 37 of 85 competing teams from universities across Egypt. Built the Smart Bathroom Safety System — an IoT-integrated safety solution combining Machine Learning, Deep Learning, and Explainable AI (XAI) to detect hazardous bathroom conditions (CO, LPG leaks, low oxygen, temperature, humidity, occupancy) in real time and trigger alerts and ventilation. Contributed across the team's pipeline — model training (Random Forest, XGBoost, LightGBM, SVM, TabNet with Optuna tuning), SHAP/LIME explainability, and the Streamlit dashboard.
Built a RAG-based Compliance AI Agent for EVA Group using Python, Pinecone, and the Groq API, orchestrated through n8n. Reached the finals among competing teams by designing a retrieval pipeline that grounded agent responses in real compliance documentation.
Credentials & Learning
Let's Connect
I'm open to internships, junior roles, freelance projects, and interesting collaborations. If you're building something with AI, ML or need a sharp creative mind — let's talk.
My Resume