Sababa Saad
AI Engineer | LLM Systems, RAG, NLP & ML | 8+ Years | Python
Ready for work
Sababa is ready for their next project!
Comprehensive Customer Career Analytics & Predictive Modeling
I performed a full end-to-end analysis on a 19 K-row customer career dataset: importing, cleaning, and handling missing demographics and company data; exploratory data analysis with distributions, correlation matrices, and striking bar charts; clustering learners into four segments to reveal training-vs-mobility patterns; and feature engineering (experience bins, training categories). I built and tuned Logistic Regression, Random Forest, Gradient Boosting, and a stacking ensemble to predict job-change (AUCs up to ~0.76).
0
18
Automated Essay Scoring using Transformer & Ensemble Models
Developed an automated essay scoring system replicating human grading using NLP and ensemble learning. Built a stacking model combining RoBERTa, DeBERTa, MiniLM, and classical ML techniques. Engineered linguistic, semantic, and readability features to enhance performance. Applied Optuna for hyperparameter tuning and evaluated results using Cohen’s Kappa for alignment with human scoring. Designed a scalable and reproducible pipeline for consistent automated grading.
Impact: Cohen’s Kappa = 0.811, reduced grading effort, scalable solution.
0
24
Multilingual NLP System for Hazard Detection
Developed a multilingual NLP system for detecting hazards and products from real-world food safety data. Processed noisy, imbalanced datasets with temporal and geographic normalization. Built features using TF-IDF, linguistic signals, and transformer embeddings. Evaluated models including Logistic Regression, SVM, XGBoost, and BERT-family architectures. Designed ensemble models optimized for competition metrics to improve generalization and robustness.
Impact: Combined Macro-F1 = 0.81, strong performance across multilingual data.
0
27
Hybrid LLM + ML Document Classification with Error Recovery
Designed a hybrid AI system for classifying domain-specific documents under noisy and highly imbalanced conditions. Combined XGBoost, transformer models (BERT-family), and retrieval-augmented LLM reasoning to improve accuracy and robustness. Introduced an “anti-breakage” mechanism to prevent LLM corrections from degrading correct predictions. Built end-to-end pipelines including preprocessing, feature engineering, model training, evaluation, and LLM fallback logic. Optimized for real-world reliability and edge-case handling.
Impact: ~89% accuracy, improved robustness, reduced manual effort.
0
35