Habiba Shawky Mohamed
Where Language Meets AI
prompt Engineer | NLP & Linguistics Specialist
View LinkedIn
Linguistic Analysis in Arabic NLP
Arabic, with its rich morphology and complex syntax, presents unique challenges in Natural Language Processing. My expertise extends to the foundational aspects of linguistic analysis, crucial for building effective NLP models for this language.
I specialize in applying advanced computational linguistics techniques for accurate morphological analysis, identifying root words and patterns, and syntactic parsing, understanding sentence structure. This enables the development of robust and nuanced Arabic language understanding systems.
Professional Summary
Habiba Shawky Mohamed is a Junior NLP Engineer with a strong academic background in Phonetics, Linguistics, and Machine Learning. She combines theoretical knowledge with practical experience in evaluating Large Language Models (LLMs) and developing various NLP systems.
Her expertise includes prompt engineering, Arabic NLP, and morpho-syntactic analysis. She is skilled in building complete ML pipelines, from data preprocessing to model deployment, using Python, Scikit-learn, and HuggingFace. Habiba is passionate about creating human-centered NLP tools that are both linguistically informed and technically sound.
Professional Experience
As a Full-time Prompt Engineer at Allendevauxx, I specialize in designing and optimizing AI interactions. My role involves crafting precise prompts and implementing robust evaluation suites to enhance LLM performance and user trust.
Prompt Design & Optimization
Designing system, user, and function-calling prompts for accurate, context-rich AI responses.
Evaluation & Quality Assurance
Building comprehensive suites (A/B, canary, red-team) to measure latency, accuracy, hallucination, and bias.
Integration & Development
Collaborating with ML Engineers to integrate prompts into RAG pipelines and developing multilingual templates.
Compliance & Monitoring
Documenting prompt versions for auditability (ISO 42001, EU AI Act) and refining for cost and relevance.
Professional Experience
AI Evaluator – Freelance
Labelbox-Alignerr (Remote) | Jan 2025 – Present
  • Focused on culturally and linguistically appropriate evaluation in Arabic NLP contexts.
  • Assessed factuality and relevance of Arabic and English responses, providing structured feedback.
AI Evaluator – Freelance
Outlier (Remote) | Sep 2024 – Jan 2025
  • Performed qualitative evaluations of LLM outputs for naturalness, logic, and task completion.
Key Projects: Constituency Parser
Habiba developed a Machine Learning-based Constituency Parser for Arabic and Multilingual Text as her Graduation Project at Alexandria University (Jan 2025 – Jul 2025).
Project Highlights:
  • Built a decision tree-based syntactic parser using prompt-engineered code and Penn Treebank-style data.
  • Achieved 86% accuracy on English and 73% on Arabic text.
  • Developed a multilingual parsing pipeline with RTL tree handling, morphological analysis, and a Flask-based interface.
  • Addressed challenges like ambiguity, reduced clauses, and Arabic word-order.
Error uploading image.
Error uploading image.
Prompt Engineering Approaches
During the Constituency Parser project, Habiba explored various prompt engineering methods to optimize chatbot performance. Four approaches proved most successful:
1
One-Shot Prompting (Desired Output)
Providing the chatbot with the desired output and asking how to achieve it. Often ineffective as the chatbot struggled to identify the core problem.
2
Pinpointing the Problem
Identifying and explaining specific problematic code parts (e.g., a function) and suggesting improvements. This approach worked better but had limitations.
3
Role Specification for the Model
Instructing the chatbot to think from a specific viewpoint (e.g., linguist or programmer).
4
Code Diagnosis (Most Successful)
Providing problematic code or output (e.g., a wrong tree) and asking for correctness assessment, then prompting for identification and fixes. This led to clearer analysis.
Independent NLP Projects
Demographic Prediction from Text Style
Mar 2025 – May 2025
  • Developed Random Forest models to predict age and gender from 681k+ blog entries.
  • Achieved 94% accuracy and R² = 0.89.
  • Engineered TF-IDF, PCA, and metadata features; applied clustering and balancing for data skew.
Arabic Morphological Analyzer
2024
  • Built a rule-based prefix-stripping analyzer using regex and linguistic heuristics.
  • Extracted Arabic stems and affixes.
  • Integrated context-aware filtering logic and word-length constraints for accuracy.
Arabic Sentiment Analysis using AraBERT
2024
  • Fine-tuned AraBERT on Arabic review datasets for binary sentiment classification.
  • Included preprocessing, tokenization, and model evaluation.
  • Achieved high precision and recall using a stratified evaluation strategy with HuggingFace Transformers.
Data Analysis Project: From Data to Insights
Habiba completed her first DEPI Data Analysis Project using only Excel, in collaboration with Ahmed Saad. They analyzed a Superstore dataset to transform raw numbers into meaningful business insights.
Workflow & Challenges:
  • Data Cleaning: Standardized column types, trimmed IDs, and formatted text using Power Query.
  • Data Modeling & KPIs: Created custom columns and measures in Power Pivot for revenue, net profit, opportunity loss, and returned orders.
  • Challenge Solved: Fixed inflated order counts due to duplicate Order IDs using a distinct count approach.
Key Insights from Superstore Analysis
Top & Lowest Profit Categories
Technology was the top category by sales, while Furniture showed the lowest profit.
Best-Selling Products
HON 5400 Series Task Chairs & Cisco TelePresence System were top sellers.
Geographic & Shipping Insights
New York City had the highest sales; Standard Class was the most profitable shipping mode.
Customer & Segment Analysis
Top 10 customers significantly contributed to revenue; Consumer segment dominated sales.
Profit & Loss Trends
Certain subcategories like Tables and Supplies recorded losses despite high sales. Seasonal sales peaks around year-end, with consistent growth from 2014–2017.
Education & Certifications
Education
  • Bachelor of Arts in Phonetics and Linguistics
    Alexandria University | 2021 – 2025
    GPA: 3.0 | Relevant Coursework: NLP, Syntax, Semantics, Machine Learning, Computational Linguistics
  • DEPI Data Analysis Track (Ongoing)
    Focus: Data cleaning, exploratory analysis, and predictive modeling using Python and real-world datasets.
Certifications
  • AI Enterprise Workflow (Coursera - Ongoing)
  • Supervised ML: Classification & Regression (Coursera)
  • Exploratory Data Analysis (Coursera)
  • Machine Learning (ITI - 2024)
  • Python Programming (ITI - 2023)
Technical Skills & Languages
Languages & Tools
  • Python, Git, Jupyter, Streamlit, Flask, Praat
Libraries & Frameworks
  • Scikit-learn, HuggingFace Transformers, SpaCy, NLTK, Pandas, NumPy, Matplotlib, Regex
NLP & ML Skills
  • Syntactic Parsing, Morphological Analysis, Prompt Engineering, Age & Gender Prediction, Tokenization, Feature Engineering, Text Classification, Model Evaluation, Arabic NLP, Decision Trees
Other Skills
  • Linguistic Analysis, Research & Documentation, Team Collaboration, Analytical Thinking
Languages
  • Arabic (Native)
  • English (Proficient)
Thank you
Made with