I buildAI assistants,trainmodelsand turndata intodecisions.

• 03Data Scientist AssistantAnalyze data & train models by chatAI & LLM+
• 04Banking Fraud Detection0.79 F1 on the fraud class across 1.42M transactionsMachine Learning+

  Detects fraudulent transactions from multiple CSV sources with inconsistent schemas and missing values, then serves predictions through a Streamlit app.

  • Cleaned and merged 1.42M transactions by engineering common keys across train/test files
  • Handled severe class imbalance with class weighting and imbalanced-learn
  • Random Forest: 99.94% accuracy, 0.90 macro F1, 0.79 F1 on the rare fraud class

  • Python/
  • pandas/
  • scikit-learn/
  • imbalanced-learn/
  • Streamlit

  Code↗

  Accuracy0.9994

  Macro F10.90

  Fraud F10.79
• 05Warranty Claims Fraud97% recall on fraudulent claimsMachine Learning+
• 06Telecom Customer Churn86% of churners caught before they leaveMachine Learning+
• 07E-commerce RFM Segmentation93,335 customers sorted into 4 loyalty tiersMachine Learning+
• 08Stock Market Forecasting40 years of AAPL forecast with ProphetMachine Learning+
• 09Sales & Profitability — Classic Models$9.6M in sales traced to its profit driversAnalytics & BI+
• 10Global Superstore SalesSales up 51%, profit lagging behindAnalytics & BI+
• 11Bank Customer Segmentation1M+ transactions → 4 product-ready segmentsAnalytics & BI+
• 12Heart Disease Risk DashboardRisk peaks at ages 50–60Analytics & BI+
• 13Netflix Content StrategyCatalog risks, mapped for executivesAnalytics & BI+
• 14Guest House & HomestayBooking site with listings & pricingWeb+
• 15Tours & Travel AgencyItineraries & destinationsWeb+
• 16Career ConsultingService site with lead captureWeb+

Also on GitHub