Real-Time Fraud Detection Pipeline
Built a streaming ML system processing 2M+ transactions/day using Kafka, Spark Structured Streaming, and XGBoost. Reduced fraud losses by 38%.
Building reliable data pipelines, predictive models, and analytics platforms that turn raw data into real business outcomes.

I'm Luxna — a data engineer and data scientist with 2+ years designing pipelines, predictive models, and analytics platforms that actually move business metrics. I care about clean architecture, honest models, and dashboards people trust.









Built a streaming ML system processing 2M+ transactions/day using Kafka, Spark Structured Streaming, and XGBoost. Reduced fraud losses by 38%.
End-to-end ML platform predicting churn with 92% precision. Deployed via MLflow on Kubernetes, integrated with Snowflake data warehouse.
Time-series forecasting engine using Prophet and LSTM ensembles. Improved inventory turnover by 24% across 400+ store locations.
Unified 12 data sources into a single source of truth. Self-serve Tableau dashboards used by C-suite for weekly business reviews.
Executed more than 30 manual test cases covering regression, smoke, functional, performance, and API testing within an Agile sprint delivery framework, reducing post-release defect rates by 5% across 3 international projects.
Developed and maintained Drupal-based web applications covering front-end and back-end systems, delivering feature enhancements and resolving defects within a 6-week project timeline.
Designed and implemented time series forecasting models using Python and machine learning (Pandas, Scikit-learn, regression ML) to predict electricity consumption across 5 years of historical data.
LLM Development
Open-source Models
Research & Notes
Conversational AI
Cloud Coding
AI App Builder
AI Image Generation
Automation
Enterprise AI
Open to consulting, full-time roles, and interesting side projects. Reach out — I usually reply within a day.