Hari Prasath Email & Phone Number

Phoenix, Arizona, United States

• Implemented Cross Validation Rules Workflow using Alteryx that aided in identifying Purchase orders (POs) that do not follow respective characteristic rules, thus decreasing the time frame required to catch invalid POs.
• Developed a Prophet model to forecast net sales at vendor and category level for vendor negotiations and estimations of future trends. The model achieved 15% reduction in forecasting error compared to previous methods.
• Employed Natural Language Processing (NLP) techniques to build Text Classifiers to extract insights from survey responses and stores google reviews, reducing the time to identify critical bottlenecks by 70%.
• Developed and implemented a DBSCAN clustering model to identify customer groups with similar purchasing habits. This analysis resulted in an 84% accurate segmentation, providing valuable insights for targeted marketing.
• Developed a predictive model using Logistic Regression to anticipate out-of-stock situations based on perpetual inventory data, thereby enhancing stock tracking accuracy by 23%
• Trained classification models to perform Inventory Classification of groceries and predict labels for a new set of product images using different algorithms namely Random Forest Classifier, SVM, Multinomial Naïve.

Tempe, Arizona, United States

• Implemented a social media analytics pipeline focused on unbiased customer sentiments using BERT Algorithm and ML libraries like pandas, scikit-learn and scipy to improve classification capabilities of customer.
• Developed Python code to perform review classification, review topic modelling, preprocessing, analysis, and postprocessing of sentimental analysis based on client-defined set of aspects, achieving an accuracy of 87%.
• Improvised the analytical set-up by identifying new set of aspects like comfort, fit and endurance limit to understand customer preferences, form meaningful marketing insights and promote appropriate set of products.

Tempe, Arizona, United States

• November 2021 - Predictive Modelling for Homesite Insurance, W.P Carey School of Business-Developed a predictive model using Homesite's database to predict whether a quoted price will lead to a purchase, resulting in.
• Implemented a social media analytics pipeline focused on unbiased customer sentiments to improve feedback on adidas products focusing on aspects in the current feedback system that necessitates improvement, thereby.
• Developed Python code to perform review classification, review topic modelling, preprocessing, analysis, and postprocessing of sentimental analysis based on client-defined set of aspects, achieving an accuracy of 87%.
• Improvised on the analytical set-up by identifying new set of aspects to understand customer preferences and form meaningful marketing insights and to promote appropriate set of products.

Chennai, Tamil Nadu, India

• Developed BI reports, dashboards, and data models from application data sources using Data Blending methods in Tableau to track key performance indicators and decode relevant patterns across datasets; achieved a 12%.
• Implemented various classification techniques to predict the fraud or fake customer applicant using LSH (Locality Sensitive Hashing) methodology based on similarity probability and achieved model accuracy of 89%.
• Built a Machine Learning Random Forest Regression model to predict rent prices across Chennai based on house characteristics and enhanced the model prediction capacity by incorporating city census data; achieved a.
• Managed a team of 4 analysts to gather, organize, and pre-process Customer feedback data using Python, SQL and Excel to aid data scientists in building marketing mix models that increased ROI by 4 basis points; honored.
• Implemented TensorFlow to build and train a deep learning model for wafer damage prediction, resulting in a 12% reduction in scrap. The model incorporated z-plot, image, and probe data to improve accuracy.
• Performed Sentimental analysis for 100K+ user reviews using Naïve Bayes approach to determine the underlying semantic factors that influence the social sentiment polarity. Data was facilitated from various sources such.

Chennai, Tamil Nadu, India

• Performed Disaster Recovery (DR) planning and server monitoring for multiple applications, including SQL and Oracle databases, achieving a record of 4 DR activities completed in 2 hours, resulting in zero downtime.
• Worked on preprocessing of ETL-based structured and unstructured high-volume datasets, thereby enabling Analytical team to work on machine readable data and improve customer conversion rate by approximately 12%.
• Devised and implemented reporting dashboards using Power BI to visualize KPIs saving 8 hours/week of manual reporting work.
• Employed high-level SQL queries to analyze structured and semi-structured data across data warehouses and lakes and deliver best price performance of high-demanded products with 15% higher data retrieval speed.
• Identified major service issues in client domains through customer data analytics, Python and SQL and provided white paper recommendations for improvements, increasing average response rates by 8%.
• Used complex SQL queries for data analysis, data visualization and for creating store procedures that enhanced real-time transaction information and buying behaviors analysis.

Chennai, Tamil Nadu, India

• Developed community structure detection model using Label Propagation Algorithm (LPA) and k-means clustering, with historical data to identify communities/clusters within a karate-club community based on common.
• Employed R to formulate the LPA algorithm, which was based on network structure to guide its progress and facilitate the model to identify certain communities of club members achieving a silhouette validation score of.
• Incorporated Deep k-means clustering and Improved Deep k-means methods to enhance the computational speed of the basic model by 10%.

Ms Business Analytics, Business Analytics

Education record