Ahmed Talaat
AI Engineer | Full Stack AI Engineer
AI Engineer and Full-Stack AI Engineer with hands-on expertise in building end-to-end intelligent systems — from raw data ingestion to production-grade model deployment. Proficient in Machine Learning, Deep Learning, NLP, Computer Vision, and MLOps. Experienced with modern cloud platforms (AWS, Azure) and frameworks including PyTorch, TensorFlow, HuggingFace, and FastAPI. Passionate about turning complex data into actionable, scalable AI products. Recognized in the NASA Space Apps Challenge and consistently delivering explainable, high-impact solutions.
- Designed and deployed end-to-end AI systems covering data pipelines, model training, REST API serving, and cloud deployment.
- Built production ML models using XGBoost, LightGBM, PyTorch, and Scikit-Learn with SHAP-based explainability layers.
- Developed NLP pipelines (sentiment analysis, classification, summarization) using HuggingFace Transformers and LSTM architectures.
- Containerized and orchestrated AI services using Docker and Kubernetes on AWS and Azure cloud environments.
- Implemented MLOps workflows with MLflow for experiment tracking, DVC for data versioning, and Airflow for pipeline automation.
- Created interactive data dashboards and Streamlit AI apps consumed by non-technical stakeholders.
- Led the development of CTRL-Sky, a geospatial weather forecasting system built for the NASA Space Apps Challenge.
- Integrated real-time satellite data feeds with a machine learning prediction pipeline deployed on AWS infrastructure.
- Engineered spatiotemporal feature extraction from multi-source meteorological datasets.
- Delivered a full-stack web interface for end-users to query and visualize weather predictions interactively.
Enterprise-grade customer churn prediction system using XGBoost with SHAP explainability, a full data preprocessing pipeline, and an interactive Streamlit dashboard for business stakeholders.
NASA Space Apps Challenge solution — a geospatial weather forecasting system ingesting satellite data and using ML models for time-series weather prediction, deployed on AWS.
End-to-end NLP sentiment analysis pipeline using LSTM networks and Transformer models (HuggingFace), with a production REST API served via FastAPI for real-time inference.