Fiham M.

Key Skills

Software

No software listed

Top Subject Matter

Generative AI for identity-preserving image generation

Cardiovascular disease risk prediction and explainable AI

Customer churn prediction classification

Top Data Types

Image

Document

Top Task Types

Text Generation

Diagnosis

Freelancer Overview

AI Image Generation Workflow using ComfyUI (ComfyUI, PuLID, Python) (2026). Core strengths include ComfyUI, PuLID, and Google Colab. Education includes Bachelor of Science (Honours), Informatics Institute of Technology (IIT) (2028) and Advanced Level, KM/Al-Ashraq M.M.V. (National School). AI-training focus includes data types such as Image, Medical, and DICOM and labeling workflows including Text Generation, Diagnosis, and Entity (NER).

Labeling Experience

AI-Based Cardiovascular Disease Risk & Recommendation System (2026)

Diagnosis

Developed a machine learning cardiovascular disease risk prediction system paired with personalized health recommendations. Applied explainable AI methods (SHAP and LIME) and graph-based similarity to support transparent and trustworthy decision-making. Structured the approach to help interpret model outputs for clinical-style risk assessment workflows. • Explainability with SHAP and LIME • Risk prediction modeling for cardiovascular disease • Graph-based similarity for supportive reasoning • Recommendation pairing with model predictions

2026 - 2026

AI Image Generation Workflow using ComfyUI (ComfyUI, PuLID, Python) (2026)

ImageText Generation

Built and deployed an identity-preserving generative AI image pipeline using ComfyUI, PuLID, and Stable Diffusion while integrating custom node workflows. Configured and optimized the AI workflow to improve stability and reproducibility by resolving dependency and GPU environment issues. Used the pipeline to generate consistent identity-preserving outputs suitable for model development and evaluation. • Identity-preserving generative image pipeline setup • Node/workflow customization for consistency • Environment dependency resolution for stable execution • Iterative runs to validate reproducible generation

2025 - 2026

Customer Churn Prediction System (2025)

Document

Created a churn prediction system using decision tree and neural network architectures while addressing imbalanced classes with SMOTE. Evaluated model performance using F1-score and AUC to support reliable classification results. Used the classification pipeline to produce interpretable churn risk outputs for business-style decision workflows. • Built churn prediction models with DT and NN • Applied SMOTE for class imbalance • Evaluated with F1-score and AUC • Produced churn risk classification outputs

2025 - 2025

Education

Work History