Pranit Jaiswal

• Delivered data-driven insights using sales data aggregated from disparate sources (SQL database and flat files) by creating an end-to-end BI solution
• Designed and implemented a star-schema data warehouse using SQL queries and ETL pipelines using SSIS to handle data for 500,000 transactions
• Visualized key performance metrics using a Tableau dashboard with interactive visualizations and provided recommendations for enhancing product sales and optimizing marketing campaign

• Analyzed ranking of reviews to make relevant information more accessible to customers using Latent Dirichlet Allocation
• Visualized the relationship between different topics within the reviews using an intertopic distance map generated through multidimensional scaling
• Recommended the integration of sentiment segmentation design with topic modeling to increase customer experience

• Performed emotional interpretation of 10,000 tweets fetched live from Twitter API and computed sentiment score of unknown terms
• Analyzed unigrams, bigrams, and trigrams to predict trending topics in different regions across the US

• Implemented and replicated spectral clustering algorithm based on “On Spectral Clustering: Analysis and an algorithm” paper
• Observed and understood the concept, advantages, and disadvantages of leveraging the algorithm in different use cases

• Designed an ER-diagram, coded and populated an original database system using SQL Server to support the operational and reporting needs of an e-commerce business
• Built multiple stored procedures with input parameters, variables, error-handling, and explicit transactions to manage INSERT activity
• Developed nested stored procedures with output parameters, business rules, and analytical functions to manage daily operational activities
• Created synthetic transaction wrappers on… Show more

• Designed an ER-diagram, coded and populated an original database system using SQL Server to support the operational and reporting needs of an e-commerce business
• Built multiple stored procedures with input parameters, variables, error-handling, and explicit transactions to manage INSERT activity
• Developed nested stored procedures with output parameters, business rules, and analytical functions to manage daily operational activities
• Created synthetic transaction wrappers on production stored procedures to validate throughput volume of 100,000+ INSERTS
• Visualized the trends in buying patterns, seller densities, regional popularity of products using Tableau dashboards Show less

• Collected large data sets with 9,800 labeled emails using Google’s Gmail API to identify trends and forecasting patterns
• Trained Logistic Regression, Multinomial and Bernoulli Naive Bayes predictive models with test accuracy of 88%
• Conducted hyperparameter tuning and evaluated performance metrics - accuracy scores, precision, recall & AUC/ROC curve

• Replicated Eigenfactor ranking algorithm using “The Eigenfactor Metrics: A Network Approach to Assessing Scholarly Journals” paper (based on PageRank) that leverage the structure of the entire citation network
• Measured the influence of articles using EigenFactor scores (eigenvector centrality and impact factor) to rank them using a weighted edge list with over 4 million relationships

• Analyzed the Airbnb listings dataset using multivariate linear regression models to develop a predictive pricing guide for homeowners based on monthly traffic
• Performed EDA and developed a pricing guide for homeowners using predictive data modeling
• Used feature engineering techniques to transform raw data into relevant features and visualized variations in prices across those features
• Developed a multivariate linear regression model to predict pricing based on 16 predictors… Show more

• Analyzed the Airbnb listings dataset using multivariate linear regression models to develop a predictive pricing guide for homeowners based on monthly traffic
• Performed EDA and developed a pricing guide for homeowners using predictive data modeling
• Used feature engineering techniques to transform raw data into relevant features and visualized variations in prices across those features
• Developed a multivariate linear regression model to predict pricing based on 16 predictors with an accuracy of 76.9% Show less

• Open Codes - Coding Culture Hackathon organized by Goethe Institut/Max Mueller Bhavan, Mumbai, India and ZKM, Karlsruhe, Germany
• Worked on the preservation of artifacts and its digital replicas using Deep Convolutional Neural Networks and Perceptual Hashing to automatically tag, compare, determine dominant colors and find relationships between images of known/unknown artifacts having a cultural bond with our society
• Monitored data quality to train predictive models (using Google… Show more

• Open Codes - Coding Culture Hackathon organized by Goethe Institut/Max Mueller Bhavan, Mumbai, India and ZKM, Karlsruhe, Germany
• Worked on the preservation of artifacts and its digital replicas using Deep Convolutional Neural Networks and Perceptual Hashing to automatically tag, compare, determine dominant colors and find relationships between images of known/unknown artifacts having a cultural bond with our society
• Monitored data quality to train predictive models (using Google Inception) on large data sets with 4,500 images curated from online data sources
• Object recognition and classification of 60,000 images, clustering of its color palettes, and hash values using metadata, TensorFlow framework, and Perceptual Hashing algorithm
• Extracted hash values of images using mathematical functions to get a unique code that describes the objects and used the code for comparing and relating attributes of the objects
• Determined dominant colors by extracting the RGB value for each image pixel, plotted it on a graph, and found the center of groups of similar pixels
• Optimized model training time from 48 minutes to 30 minutes with an accuracy of 70\% and visualized image classifiers and relationships using the Neo4j graphical database
• Secured the 1st place in the Hackathon and was invited to the Annual Goulash Programming Night (GPN) Conference held in Germany for a feature presentation Show less

• Senior Year Capstone Project at Fr. C. Rodrigues Institute of Technology under the University of Mumbai curriculum
• Researched the collaboration of the Bayesian Linear Hierarchical model and Google’s Wide and Deep Learning models to avoid cold start problem in a data-driven approach for recommender systems on mobile devices
• Developed a recommender system for an Android Launcher product design to analyze data metrics like user preference, location and time of the day for forecasting… Show more

• Senior Year Capstone Project at Fr. C. Rodrigues Institute of Technology under the University of Mumbai curriculum
• Researched the collaboration of the Bayesian Linear Hierarchical model and Google’s Wide and Deep Learning models to avoid cold start problem in a data-driven approach for recommender systems on mobile devices
• Developed a recommender system for an Android Launcher product design to analyze data metrics like user preference, location and time of the day for forecasting user-centric tasks to cater to personalized mobile user experience
• Other features of the launcher:
1. Search apps, contacts, settings toggle, Google, Wikipedia from the home screen of Android device
2. Favorites bar for quick access to apps
3. Blue light filter
4. Password protection
5. Hiding installed applications from the app drawer
6. Support for third-party icon pack and 6 different themes
• Secured 1st place in the Senior Year Graduate Project Competition amongst 19 teams.
• Tools and Libraries: Android Studio, Android SDK, Kiss Launcher
• Programming Languages: Python, Java, XML, SQLite Show less

• Smart India Hackathon - world's most prominent Hackathon organized by the Government of India
• Led a team of five engineers and conceptualized an end-to-end product development life cycle for de-duplication of large data sets having one billion long integers with a scalable approach for the future needs of the clients
• Programmed sliding window algorithm with blocking technique and modeled data-parallel processing pipelines on Hadoop MapReduce
• Optimized time complexity from… Show more

• Smart India Hackathon - world's most prominent Hackathon organized by the Government of India
• Led a team of five engineers and conceptualized an end-to-end product development life cycle for de-duplication of large data sets having one billion long integers with a scalable approach for the future needs of the clients
• Programmed sliding window algorithm with blocking technique and modeled data-parallel processing pipelines on Hadoop MapReduce
• Optimized time complexity from O(n^2) to O(n) + O(nlogn) with a performance increase of 67.4% by parallelizing it on a Hadoop cluster with 1 master and 3 slaves
• Improved efficiency in execution time from 43 minutes (single node) to 14 minutes (multi-node) approximately
• Qualified for final round among more than 40,000 participants
• Transmission media: Netgear Switch with 100mbps transfer rate Show less