A Power BI HR Analytics Dashboard built on the IBM Employee Attrition dataset. The project uses Power Query, DAX, Tabular Editor, and DAX Studio to analyze employee turnover, salary impact, overtime, satisfaction, and attrition risk across 1,470 employees through four interactive dashboard pages.

Built an end-to-end retail analytics pipeline using Python, PostgreSQL, SQL, and machine learning to analyze 9,994 sales transactions. Developed automated data cleaning, advanced analytics, forecasting models, and interactive dashboards (Tableau, Power BI, Plotly, Streamlit), generating actionable business insights and recommendations that identified opportunities to improve profitability by up to 68%.

Data Analyst with experience in Excel, Power BI, Tableau, Orange Data Mining, and MongoDB. My portfolio showcases projects in data cleaning, data modeling, dashboard development, machine learning, and NoSQL database design, transforming complex datasets into clear, actionable insights across domains such as healthcare, sports, demographics, sales, and e-commerce.

A vanilla JavaScript country explorer app with interactive maps, rich country data, and AI-generated insights for a simple and visual way to discover the world.

A complete end-to-end deep learning pipeline for vehicle detection in SAR imagery using YOLOv8-OBB on the SIVED dataset. The project covers data processing, exploratory analysis, oriented bounding box handling, model fine-tuning, evaluation, and deployment via a Gradio web app, achieving near state-of-the-art performance in a challenging radar-based detection setting.

RehanPulse is a real-time developer dashboard that centralizes activity from tools like GitHub, Vercel, and Firebase into a single interface. It delivers live updates on commits, CI pipelines, deployments, and alerts using Server-Sent Events and Firestore listeners, removing the need to switch between multiple tabs, and is built with modern technologies like Next.js and TypeScript.

Discussing building FinanceTracker, a secure personal finance app. Highlighting key choices in technology, authentication, and data security, showing how careful design ensures usability, privacy, and reliability for managing personal finances.

A compact, end-to-end guide to fine-tuning Qwen2.5-VL, a vision-language model, on astronomy-specific data using Unsloth, covering the full workflow from dataset preparation to training, with a focus on efficiency and minimal resource usage.

Fine-tuning Meta's Llama 3.2 11B Vision model on a 250-image astronomy dataset using Unsloth and LoRA, completing training in under 12 minutes on a free Tesla T4 GPU, producing a model that accurately describes astronomical images and is deployed via Gradio on Hugging Face.

A deep learning pipeline using transfer learning to classify chest X-rays into COVID-19, normal, viral pneumonia, and lung opacity. Includes data exploration, cleaning, augmentation, model training, evaluation with metrics like confusion matrices and ROC curves, and interpretability via Grad-CAM visualizations.

Comparing custom CNNs, Hugging Face vision transformers, and Xception transfer learning for vehicle classification, showing that transfer learning - especially Xception with two-phase fine-tuning - greatly outperforms training from scratch on small datasets.

A practical deep learning journey using transfer learning with InceptionV3 to classify sports balls. This project covers data cleaning, class balancing, preprocessing pipelines, two-stage training, and rigorous evaluation - highlighting why data quality and smart engineering matter more than model complexity.

High-accuracy InceptionV3 pipeline for cat vs. dog image classification with robust preprocessing.

End-to-end deep learning pipeline using InceptionV3 and Grad-CAM for accurate and interpretable brain tumor MRI classification.

Built an end-to-end audio classification pipeline using CNNs on Mel spectrograms, with data cleaning, augmentation, and deep learning for reliable multi-class predictions.

Built an animal image classification pipeline using TensorFlow and InceptionV3, focusing on thorough data cleaning, preprocessing, and transfer learning for accurate results.

A practical walkthrough of how I built and trained a deep-learning model to denoise images and boost classification performance.