AI/ML Engineer at EPSILON

Designed and productionized end-to-end machine learning and Generative AI workflows using Python, SQL, Airflow and cloud data platforms, covering ingestion, feature engineering, training, validation, batch scoring and monitoring. Built LLM-powered customer intelligence solutions using LangChain, prompt engineering, contextual retrieval, embeddings and NLP to generate personalized recommendations and insights. Developed and optimized 20+ Airflow DAGs to automate ML/data workflows and reduce recurring manual intervention by ~35%. Engineered classification/predictive models (propensity, churn, segmentation) improving performance by ~15–20% vs baseline. Established reusable preprocessing/feature engineering/validation and model evaluation components, reducing development turnaround time by ~30%. Implemented model performance monitoring, data-quality checks and production validation to detect degradation and upstream data issues early. Built scalable Python inference components and APIs,

Machine Learning Engineer at TRIGENT SOFTWARE

Developed ML solutions in Python/SQL using scikit-learn and TensorFlow for classification, forecasting, customer analytics and anomaly detection. Built reusable preprocessing and feature-engineering pipelines for structured datasets with millions of records. Designed and evaluated classification/regression models (Logistic Regression, Random Forest, Gradient Boosting, XGBoost) improving selected accuracy by ~18% vs baseline. Developed LSTM and ARIMA forecasting models using seasonality, lag/rolling features and external variables, improving forecasting accuracy by ~27% vs baseline seasonal methods. Automated model-training and data-processing workflows, reducing manual effort by ~40% and improving reproducibility. Performed EDA, outlier handling, imbalance handling and hyperparameter tuning; built evaluation frameworks using cross-validation and metrics such as ROC-AUC, RMSE and MAE. Integrated models into downstream applications and data workflows, supported testing/validation/deploym