Debangsha S

Freelance Machine Learning Engineer

Freelance

Jan 2015 - Present • 11 yrs 4 mos

1. I developed a pipeline for an online retailer's customer behavior prediction system. I used Apache Spark, MLlib, Python, and Kafka for the project. I used Spark's MLlib to implement logistic regression and decision tree models to complete the task. After finishing the project, the company improved its targeted marketing strategies, leading to a 15% increase in sales. 2. I worked with an online retailer to classify and categorize their products on their website from images. I used Apache Spark, TensorFlow, Python, and Convolutional Neural Networks (CNN) for this project. I used the Spark-TensorFlow connector for scalable training. I achieved a 93% accuracy on the ML model on the test dataset they provided me. The retailer used my ML pipeline for inventory management. 3. I created a real-time fraud detection pipeline for a neo-banking startup company based in Canada. I used Apache Spark Streaming, MLlib, and Python to implement a Random Forest model in MLlib for anomaly detection. After implementing the ML pipeline, the company reported a reduction in fraudulent transactions by 20% and improved response time. 4. I completed a project on sentiment analysis for customer feedback for an online travel business. I used MLlib, Python, and NLP libraries (NLTK, SpaCy) for this project. I trained a support vector machine (SVM) model that was fine-tuned with cross-validation on the online reviews. The final product enhanced the understanding of customer satisfaction leading to improved products and services. 5. I worked with a local hotel chain for their social media sentiment analysis, which they used for better customer service. I streamed social media data using AWS Kinesis and then processed it in real-time with Spark Streaming. I used AWS Lambda for preprocessing and NLTK and SpaCy for sentiment analysis. I stored results in DynamoDB for quick retrieval and analysis.