Muneef Muhammed

Detail-oriented Automation Engineer and Data Science graduate student with two years of experience in test automation, specializing in system debugging and log validation with Robot Framework and Selenium. Skilled in Python web development and data engineering automation, focusing on AI-driven solutions. Actively pursuing a Master of Data Science at UWA, strengthening expertise in data analysis and machine learning.

Natural   Language Inference System for Sentence Relationship Classification
Python, PyTorch, NLP, Deep Learning

• Built deep learning models to classify relationships between premise-hypothesis sentence pairs using BiLSTM, ESIM-lite (biGRU with cross-attention), and Transformer architectures.
• Developed a text preprocessing pipeline including normalization, tokenization, and dataset preparation for noisy web-based text data.
• Conducted hyperparameter tuning and ablation experiments to analyze the effect of attention mechanisms on model performance.
• Generated attention heatmaps and visualizations to interpret model behavior and understand semantic relationships between sentences.

AI   Guardrail System for Automated Threat Modelling Pipelines
Python, FastAPI, JSON Schema, LangGraph, Neo4j

• Designed and implemented a guardrail service to validate structured system descriptions before automated security threat modelling analysis.
• Developed a FastAPI microservice with endpoints for prompt refinement and JSON validation to ensure structured and reliable inputs.
• Defined a canonical JSON schema for Data Flow Diagrams (DFDs) and implemented rule-based validation checks including unique ID enforcement and security constraints across trust boundaries.
• Integrated the validation layer with a LangGraph-based workflow and Neo4j graph database used for STRIDE threat modelling analysis.
• Built a benchmarking script to evaluate system acceptance rate and latency with deterministic validation and optional LLM-assisted prompt refinement.

Machine Learning Classification   & Face Recognition
Python, Scikit-learn, PCA, Machine Learning

• Built and evaluated classification models including Logistic Regression, Random Forest, and Support Vector Machine using the Labeled Faces in the Wild (LFW) dataset.
• Applied Principal Component Analysis (PCA) to reduce high-dimensional facial image data into a compact set of principal components.
• Implemented an ensemble Voting Classifier to combine multiple models and improve classification performance.
• Evaluated models using accuracy, precision, recall, and confusion matrices to compare model effectiveness.
• Created visualizations including eigenfaces, confusion matrices, decision boundaries, and feature importance heatmaps to interpret model behavior.