Tushar Pandit

- Working on designing and implementing data pipeline using Amazon Appflow to sync data from salesforce to datalake singlehandedly
- Working on creating the appflow resources dynamically using terraform; constantly in touch with the aws-terraform maintainers for bug fixes; contributed to a few issues to expedite the fbug fixing effort
- Working on creating a spark application for creating delta tables which runs periodically using Apache Airflow
- Working on creating another spark job… Show more

- Working on designing and implementing data pipeline using Amazon Appflow to sync data from salesforce to datalake singlehandedly
- Working on creating the appflow resources dynamically using terraform; constantly in touch with the aws-terraform maintainers for bug fixes; contributed to a few issues to expedite the fbug fixing effort
- Working on creating a spark application for creating delta tables which runs periodically using Apache Airflow
- Working on creating another spark job to validate the data sync
Technologies used: Amazon Appflow, Databricks Delta tables, AWS S3, AWS Glue, AWS Athena. Apache AIrflow, Apache Spark, Scala, Hashicorp Terraform Show less

- Worked on data migration to move smb datasets from redshift to data lake
- Worked on designing and implementing a backfill app to backfill historical data
- Tools used: Apache spark, kubernetes, prometheus, grafana, scala, airflow

- Worked on onboarding sales ML models to newer infrastructure at ZipRecruiter
- Successfully used Weights & Biasis (known as wandb) to store model artifacts
- Deployment of these models to k8 namespaces using an internal service called runway (to track model lifecycle)
- Helped the Data Scientists by onboarding to models to jupyterHub for easy training and experimentation
- Tools used: wandb, runway, flyte, kubernetes, python, airflow

- Worked as a software development engineer for the Business Web Services team in Amazon Business • Worked on exposing amazon business reconciliation APIs to external clients and customers
- Worked on Internationalization of our services to launch it in new regions/ marketplaces
- Worked on implementing Integration Test and Canary Test frameworks
- Technologies: AWS, Cloudwatch, Api Gateway, IAM, DynamoDB, S3, Cloud Formation, CDK, Lambda, VPC, Java, REST, Guice

- Working on the Visa AI Platform Team
- Building an AI Platform to be leveraged by the various Data Science Teams within Visa
- Working on Feature Toolkit wherein data engineered the streaming and batch data which is inserted in a dataset called DRUID
- Worked on creating a Microservice in Scala which queries DRUID, computes aggregates which are further pushed to KAFKA and consumed by a receiving system
- Worked on modifying Ansible scripts to make HashiCorp Consul available in… Show more

- Working on the Visa AI Platform Team
- Building an AI Platform to be leveraged by the various Data Science Teams within Visa
- Working on Feature Toolkit wherein data engineered the streaming and batch data which is inserted in a dataset called DRUID
- Worked on creating a Microservice in Scala which queries DRUID, computes aggregates which are further pushed to KAFKA and consumed by a receiving system
- Worked on modifying Ansible scripts to make HashiCorp Consul available in different data centers for Disaster Recovery
- Successfully implemented test automation using Robot Framework.
- Technologies used: Spring Boot, SCALA, ProtoBuffs, GRPC, YAML, HashiCorp Consul, Hadoop, KAFKA, DRUID, Avro, Microservices, Robot Framework, Ansible Show less

- Workied as a full-stack software engineer with concentration in backend in the Fraud and Auth domain
- The product is used by the issuing banks to define rules and detect fraud on any transaction
- Worked on implementing data loader in order to support jobs in the legacy system
- Technologies used: Spring MVC, Angular 5, Linux, REST APIs, Bitbucket, Agile SCRUM, JIRA

- Successfully built an audio detector which takes a fragment of audio as input and determines whether that fragment contains the sound of gunshot or not.
- Input was a set of audio files which was used for training
- Extracted audio features (crossing rate and spectral centroid) using Librosa
- Associated each feature vector with a class label
- Trained the model using SVM with RBF kernel, K-Nearest Neighbors, Gaussian Naive Bayes and Neural Nets
- Performed 10-fold… Show more

- Successfully built an audio detector which takes a fragment of audio as input and determines whether that fragment contains the sound of gunshot or not.
- Input was a set of audio files which was used for training
- Extracted audio features (crossing rate and spectral centroid) using Librosa
- Associated each feature vector with a class label
- Trained the model using SVM with RBF kernel, K-Nearest Neighbors, Gaussian Naive Bayes and Neural Nets
- Performed 10-fold cross-validation to compute the accuracy and made predictions on the unseen data
- Computed performance metrics like accuracy, precision, recall, f1-score. Neural Nets performed better.
- Tools used - iPython Notebook, python 2.7, Librosa, scikit-learn, pandas, numpy Show less

- Researching the data using Spotify’s APIs which would help in predicting danceability of a song
- Performed data pre-processing (data imputation, normalization, standardization) and Exploratory Data Analysis (EDA) by plotting Scatter Plot and HeatMap. Implemented Feature scaling and Feature selection
- Successfully implemented Logistic Regression, Gaussian Naïve Bayes, Support Vector Machines with Linear, RBF and Polynomial Kernels, Neural Networks, K-nearest neighbor algorithms with… Show more

- Researching the data using Spotify’s APIs which would help in predicting danceability of a song
- Performed data pre-processing (data imputation, normalization, standardization) and Exploratory Data Analysis (EDA) by plotting Scatter Plot and HeatMap. Implemented Feature scaling and Feature selection
- Successfully implemented Logistic Regression, Gaussian Naïve Bayes, Support Vector Machines with Linear, RBF and Polynomial Kernels, Neural Networks, K-nearest neighbor algorithms with and without Principle Component Analysis (PCA)
- Computed accuracy, precision, recall and f-score by plotting the classification report to decide the optimum algorithm
- Libraries used: sklearn, seaborn, pandas, numpy, matplotlib, etc.
Show less

- The project was to estimate the variables which impact the bank churn
- Connecting Tableau to an excel file
- Visualizing an Ad-Hoc A-B Test
- Creating Aliases and adding a reference line
- Finding anomalies using Ad-Hoc A-B tests and validating the data/approach
- Creating and using Bins to convert numerical data(measures) to categorical data(dimensions)
- Visualizing distributions and creating A-B tests for numeric vars
- Combining two charts in Tableau, comparing and… Show more

- The project was to estimate the variables which impact the bank churn
- Connecting Tableau to an excel file
- Visualizing an Ad-Hoc A-B Test
- Creating Aliases and adding a reference line
- Finding anomalies using Ad-Hoc A-B tests and validating the data/approach
- Creating and using Bins to convert numerical data(measures) to categorical data(dimensions)
- Visualizing distributions and creating A-B tests for numeric vars
- Combining two charts in Tableau, comparing and drawing conclusions
- Validate the estimates by running Chi-Squared Tests Show less

- Read the dataset in MATLAB and performed random shuffling of data
- Performed data pre-processing and spilt the data into training and testing set
- Performed 5-fold cross-validation and implemented KNN (k values ranging from 1 to 17) algorithm by computing distance matrix
- Finally, computed average accuracy for every nn on the on all 5-folds
- Plotted a graph of k (from knn) versus accuracy to find the best k.

The purpose of the project was to determine the probability of a person having malignant or benign cancer. The dataset used was "https://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+(Diagnostic)"
The project included following steps:
- Reading the dataset into panda's data-frame
- Separate the target and the independent variables
- Transform the class labels to integer values
- Divide the dataset into training and test values
- Create a pipeline to perform chain of… Show more

The purpose of the project was to determine the probability of a person having malignant or benign cancer. The dataset used was "https://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+(Diagnostic)"
The project included following steps:
- Reading the dataset into panda's data-frame
- Separate the target and the independent variables
- Transform the class labels to integer values
- Divide the dataset into training and test values
- Create a pipeline to perform chain of operations like standardizing the data, dimensionality reduction using PCA, and then applying logistic regression algorithm
- Calculating the accuracy and f-score, printing the confusion matrix
- Applying grid search approach and tracking the performance
- Implemented the algorithm using pandas and scikit-learn
Show less

- Performed geospatial database operations on large datasets stored in distributed systems using Hadoop, Apache Spark, Scala, GeoSpark library in Linux
- Performed cluster analysis (efficiency, memory usage and CPU usage for each node) using Ganglia
- Successfully implemented an algorithm for Spatial-Temporal hotspot analysis that included determining the top 50 hotspots for taxi pickups in New York city in January 2015 using Getis- Ord statistics

A database was provided which consisted of housing data. The objective of the project was to determine the house prices for new data and also understand regression. The dataset used was 'https://archive.ics.uci.edu/ml/machine-learning-databases/housing/housing.data'. It involved the following steps:
- Reading data into panda's data-frame
- Visualizing it's characteristics using Exploratory Data Analysis(EDA)
- Creating a correlation matrix and plotting a heat-map
- Separating the… Show more

A database was provided which consisted of housing data. The objective of the project was to determine the house prices for new data and also understand regression. The dataset used was 'https://archive.ics.uci.edu/ml/machine-learning-databases/housing/housing.data'. It involved the following steps:
- Reading data into panda's data-frame
- Visualizing it's characteristics using Exploratory Data Analysis(EDA)
- Creating a correlation matrix and plotting a heat-map
- Separating the dependent and the independent variables
- Scaling the values using Standard Scalar
- Building the linear regression model and calculating the performance metrics like R2-score and MSE.
- Applying multiple regression, ridge regression, lasso regression and random forest regression and tracking the performance of each model.
- Implemented the algorithm using pandas and scikit-learn Show less

• Developed an independent system to track personal medicinal needs.
• Followed Personal Software Process (PSP0) for software development
• Documented PSP documents such as Defect Log, Project Plan Summary, PSP Retrospective, Time Log and also other documents like Concept Proposal Document, Design Document, Design Manual, Software Requirement Specification
• The project was developed using JAVA Swing and ANT tool was used to create an executable JAR

• Worked on Persistent Systems’ in-house project which was an independent test tool to be used for recruitment.
• Included registration of new candidates (personal, education, family information). Worked on the UI independently along with writing APIs and Stored Procedures. Successfully provided live support for the project with very limited guidance.
• Based on a test-pin which was generated using Microsoft’s ERP tool Axapta.

• Worked on a client project which was in Asp.NET MVC which was to gamify Persistent Systems’ recruitment process.
• Building missions and corresponding tasks in missions; awarding the users virtually through points and badges on successful completion of those tasks and missions. It had two portals namely admin and user. Successfully handled the user side independently.
• Skills: ASP.NET MVC, Web API, HTML, CSS, AJAX, JQuery, Bootstrap

• Worked briefly on a client project in ASP.Net Entity Framework which was based on taking personality surveys.
• My task was to upgrade the UI as per the requirements, resolve bugs and get the code reviewed before checking it in the Source Control.
• Skills: ASP.NET Entity Framework (Code-First approach), HTML, CSS, MS SQL

• Automation of company’s document generation system for their employees.
• Successfully deployed the project and is used by about 100 employees per month in the company today.
• Generation of PDF in ASP.Net which is used for generating approximately 100 pdf documents every month.
• Accessing the company’s Active Directory to retrieve employee database using Lightweight Directory Access protocol(LDAP).
• Skills used: ASP.NET C#, HTML, CSS.

• A desktop application used for computation of bill in electronic store and maintained records of the customers, the employees in the etc.
• Object oriented programming concepts like inheritance, abstraction, polymorphism, etc. were implemented efficiently.
Skills: JAVA, MS Access.

• A desktop application which provided instant medicine lists for a particular disease and also identify the disease based on symptoms.
• An individual project in which concepts of classes, inheritance, file handling, polymorphism, etc. were successfully implemented.
• Skills: C++

• A desktop application that maintained details of employees, stocks, customers as well as generated receipts for them.
• The concepts of normalization, querying, connectivity to databases were implemented in this project. Handled the UI and the connectivity to the database along with the documentation independently.
• Skills: Visual Basic, Oracle.