Arun kumar Jegarkal

Master of Science in Computer Science

Bachelor of Engineering in Computer Science

Diploma in Computer Science

Offline no shortage priority Analysis (Peterbilt)

Management spends time using EWI data to lookup defects and shortages on trucks to create priority lists to guide the offline flow to park trucks, the process of creating these lists is time consuming and the lists are not up to date by the time they reach the operators.
* I built a dashboard that will help the management to quickly see the number of defects on the trucks.
* Group and drill down to the trucks by part shortage category, line location , defect owner
* Identify and Prioritize low defect trucks and work on those so that those are ready to deliver
* See the truck parking location in the lot so that drivers can bring it to the plant for the rework.
* As this dashboard is real time which eliminates the manual work which is done before every shift.

Materials Parts Summary Analysis (Peterbilt)

This analysis was for the materials team to identify the number of trucks with part shortages by build day, and allows the plant to adjust build rates to reduce the impact of part shortages.
Key KPI were:
* Ability to see the Net on hand parts quantity which includes on hand and shipped quantity.
* Ability to see part utilization for the next 30 business days.
* Visibility to identity the part shortage using utilization calculated using build rate, required parts quantity based on scheduled build date/ chassis line order number
* Ability to see the shortage across all the manufacturing plants
* Ability to see chassis that are affected by part shortage, drill through part family

Truck Traffic Analysis (Peterbilt)

Project Overview & Objectives Dealer Development needs to implement the connected truck data into various analyses. Using the connected truck data, Dealer Development can accurately understand the level of truck activity in a given dealership location’s territory. This insight produces more precise recommendations for the need of more (or possibly less) of the following in a given dealership’s area/market:
* Number of dealerships
* Number of Service Bays
* Hours of Parts & Service operation
* Number of Mobile Service units
* Number of Outside Parts Salespeople

In this Project I worked as a Data engineer and performed below duties:

• I conducted the POC to prove that processing millions of remote diagnostics data in python and finding the respected dealers using the user provided json territory files.
  
• I developed the architecture to run the dealercd assignments on aws environment and using snowflake remote diagnostics and dealer territory files and store the results back to snowflake. And to implement business requirements.
  
• I also did the cost estimation of the project which mainly includes AWS Services cost and Snowflake compute and storage cost.
  
• Developed terraform templates for deploying the aws infrastructure on dev, test and prod accounts. Which involves AWS EC2, S3 bucket, Lambda function, SQS, SNS email
  
• Performance of the process was a key part of the project so used python multiprocessing based on EC2 instance cores and daemon thread to sync logs to s3 while executing the script.
  
• Developed Jenkins pipeline for deploying the infrastructure. And pipeline to build docker containers using the code from VSTS repo and deploy an EC2 instance and also pipeline to execute docker containers when file is uploaded to s3 and also on cron schedule.
  

Engine Reliability Analysis (Peterbilt)

Project Overview & Objectives The Engine Reliability Analysis (ERA) project was initiated as a directive from PACCAR's executive team. The MX13 EPA17 engine family is experiencing multiple catastrophic failure modes which have resulted in a significant increase in warranty costs. These issues also have a profound impact on the customer downtime, and the customer's perception of PACCAR's quality.
The ERA project is assisting with two primary use cases:
1.) Relative Risk of Failure - For each failure mode and each affected chassis, provide a projected probability of failure over the period of interest (e.g. next 12 months). Ranking by this probability provides a relative risk that allows the business to prioritize resources (dealership capacity, parts, warranty dollars, etc)

2.) Warranty Forecast - Projecting the number of failures that will occur within the warranty period. These projections are used to set the accrual amount for PACCAR's financial reporting. In addition to standard reliability projections, simulations are used to estimate the impact of the planned field interventions (campaigns, customer education, etc).

3.) Root Cause Analysis - Provide engineering insights with variables that are highly correlated with reliability to assist with generating root cause hypothesis. It is important to note that the analytics findings provide correlated variables, but these are not necessarily causal factors.

In this Project I worked as a Data engineer and performed below duties:

• Designed the architecture running ML model on AWS using services like S3, SQS, SNS, EC2, ECR, Lambda and data from snowflake and parquet files
  
• Designed snowflake er diagram for storing ML data, artifacts, build logs, database views.
  
• Created the snowflake views using multiple tables/views from multiple databases for pulling the chassis, ops, EASOP and cab database and summarizing the remote diagnostics records for latest miles, engine revolution and engine hours.
  
• Developed the terraform templates for deploying the aws infrastructure on dev, test and prod accounts.
  
• Developed multiple Jenkins pipeline for deploying infrastructure, pipeline for deploying ML code and pipeline for running the models when file placed on S3 bucket.
  

Warranty Repairs Per one Hundred (R100) (Peterbilt)

Responsibilities:

• I wrote a complex snowflake queries and views to calculate R100 for NIS/3MIS, Vehicle and Engine meeting at ATA and Parts level using Warranty, OPS and CAB database for the calculation R100 = total failures / total parts used.
  
• Python scripts are written to read data from snowflake db and calculate R100 and store the results into project specific schemas.
  
• Designed an architecture for running a script on schedule using AWS services (EC2, SNS, CloudWatch Event, S3, ECR)
  
• Developed terraform template for deploying the AWS Infrastructure.
  
• Developed Jenkins pipeline for running the deployments and running the scripts on schedule using cron job.
  
• Used VSTS as a code repository and Jira for project management.
  
• Developed a tableau dashboard for visualizing the R100 at ATA and Parts level, used various features of tableaus like charts, filters, actions, show hide menu, cross charts filtering, drill through.
  

Failure Mode Identification Project (Peterbilt)

I handled the data engineering apart of this project
Responsibilities:

• Setting up the snowflake schemas, tables and views
  
• Developed architecture for deploying the pipelines
  
• Wrote python data util scripts for data scientists for reading and writing data to snowflake.
  
• Developed flask application:
  * for downloading the claims and tagged failure mode for warranty meeting
  * for uploading validated files to snowflake by performing validation check
  * uploading and downloading Top ATA/Parts status files, Golden Set, Training data.
  
• Deployed flask application on AWS EC2 instance using docker container.
  
• Sync flask application log files to AWS S3 using APScheduler package.
  
• Developed terraform template for deploying AWS infrastructure
  
• Used VSTS as code repository and Jira for project management
  

Netrush Platform

This is a Netrush in-house retail domain project, focused on developing e-commerce platform. I am responsible for gathering user requirement and translating into development features. Also involve in testing tools.

Netrush Portal

This is a Netrush Analytical and reporting project, focused on developing reports and dashboards to get insights of sales, inventory, compliance, violations, pricing etc.,.

National Eligibility Database (NEDB Carrier Loads)

This is Insurance domain project for the client HMS (Healthcare Management System). We receive files on timely bases from Commercial Insurance carriers, Third party, Pharmacy Benefit Managers, Government Payers for which we do ETL (Extract Load and Transformation) and perform Data match using proprietary logic in order to Cost Avoidance and Recovery.

I was responsible for the Extraction, Transformation and Loading of data from multiple sources using Informatica ETL tools and SQL. To perform these duties, a deep understanding of the Data Warehousing concepts were required. Thus, I used Informatica Designer for designing and developing mappings from varied transformation logic using Expression, Source Qualifier, Filter, Router, Joiner, Lookup, Sorter, Aggregator and Update Strategy. I also designed ETL logic for Dimension Tables using SCD – Type1 and Type2. Last, I worked with heterogeneous data sources like DB2, Flat Files and Mainframe Files to accomplish the goals of the organization.

• Preparing Minutes Of Meeting
  
• Giving internal training on Informatica PowerCenter.
  
• Preparing Informatica training materials.
  
• Updating the Track.

Geospatial Visualization of Twitter Data

Introduction

This is an National Science Foundation funded research, it is focused on building a model for the researcher to collect tweets using Twitter API and analyze the trends in Twitter across Texas counties using hashtags to find the topics people are discussing every day. As hashtags are metadata, they help to recognize topic discussion at a first glance. Analyzing the evolution of topics over time shows how a discussion increases or fades away when new topics emerge. Discussions can also be suddenly switched to new topics. Human discussion behaviors can be studied through this type of analysis of topic evolution. Part of this analysis is frequency distribution of the hashtags which is calculated to highlight the use of the hashtags for each day. Another part of this analysis is applying topic modeling algorithms to tweets to bring out the generalized view point of the tweets. Over all tweets, conclusions are difficult to derive as many tweets consist of diverse topics; thus, topic analysis methods applied on a smaller set of tweets pertaining to hashtags shows more uniformity of thought in the tweets. These topics and hashtags are mapped to their respective county on a daily and weekly basis to find the co-occurring and identical discussion across the state of Texas. Further, to know the opinion of tweeters on these topics, tweets related to the hashtags are extracted and the sentiment, positive or negative, of each tweet is calculated using supervised machine learning algorithms. These positive and negative sentiment tweets are mapped to their respective county to find tweeter perspectives on the topics at each county level; thus, enabling researchers to factor in tweet analysis to their geospatial research of disease indicators, governmental policy issues, and social norms.

Application was built in Python using different packages like for UI {wxPython}, for charts and reports {bokeh}, for Analysis {pandas, SciKit-Learn, NumPy, SciPy, Plotly, NLTK, Gensim, Tweepy,json,urllib etc.}.

Movie Scope

Abstract

We implement a system through which we can analyze and predict the box office collection of the movies. We build the system in such a way that people from various backgrounds can also interact with the system and the data we have. Through the interaction they can also find hidden trends and insights which some of us miss. The data contains various variables which are compiled from various sources. The data can be extremely exhaustive but we have tried to collect the information essential for the analysis of trends and insights.

Introduction

Each year there are hundreds of movies which are released and being produced. The revenue of the movies depends on various factors like genre of the movie, critics review, quality of trailer, stardom of lead actor, studio of the movie being produced, director of the movie and various people included in movie making process. It can also depend on the current state of the economy, spending trends and habit of people, stock market and various other economic factors. The movie industry was worth more than 100 billion dollars in 2016[1]. And it is only expected to grow with number of new movies and stories and also the growing population. So it becomes useful to predict how much amount of money will one movie generate based on historic figures. In this paper we take into amount all the factors described above like genre, critics review, lead actor, studio etc. This demo helps to analyze and visualize the trends of the box office collection. End user can visualize the collections over the years based one the MPAA (Motion Picture Association of America) Rating's. Classify based on years, title name. Word cloud of top two movies help the user to know more about the movie summery, cast, genres. Review sentiment helps to know people opinion about the movie.

A High Performance, Scalable Event-Bus System for Distributed Data Synchronization (Java Application)

Introduction

In this project, we basically focus to overcome the limitations of the existing techniques and our primary goal is to provide the support for fully transactional communication between the clients and the eventbus thereby making the communication more effective. Websockets replaces the native socket and multithreading concepts of java, which is the efficient way of establishing the pub-sub connection. Serialized the message delivery between publisher-eventbus-subscriber by using ‘Avro’ which is one of the serialization techniques used in hadoop.

Gnutella style P2P file sharing system (Java Application)

Introduction

This project is about implementation of Gnutella-style peer-to-peer (P2P) file sharing system, where signal peer act as both client and server with performing the file search within the specified network without using the Indexing server. I have implemented using java socket programming and multithreading concepts. When peer is started its respected server thread is started which runs in the background, given config file is read to find its peer and starts the client thread for each neighbor and when the neighboring peer gets online connection is established this connection is used to search file among the network. Server thread is used to search file within its local directory and store the result into the list once the local search is performed it reads the config file to start the client thread to all its neighbors. time-to-live (TTL) is attached to the message along with the filename need to be searched. TTL is used for preventing query messages from being forwarded infinitely many times. Client thread is used to send the file to search to servers which are connected to it. Server Download thread performs the downloading of file from the peer which is selected by the user. Same server and client threads are used for broadcasting the file modification message to all the neighboring peers which has that modified file.

Napster style peer-to-peer file sharing system(Java Application)

Introduction

I have used Socket Implementation of Napster-style peer-to-peer (P2P) file sharing system. The main idea of this project is to Create the indexing server which stores all the file details and bringing it online so that all the peers connect to it and register its files on the server. Once files are registered it provided with 2 option to search for the file in other peer or can exit the execution, if search is selected file name is asked and port number and sent to the server, sever will return the peer id’s which as the respected file and then client has to select the peer id from which file to be downloaded. Downloaded file will be stored into the requested peer’s shared directory. Every Peer can provide its shared director and can register its files so that other peer’s can search and download them. Every Client Server connection is done using Sockets, given the port number and the Server name client establishes a connection with the Server registering files and searching files among peers.

Visualizing time series data

Introduction

Representing the unemployment information for different states in USA based on the data set available Bureau of Labor and Statistics . This website allows the user to select the date range to generate a time series visualization.

Functionalities

• User can zoom into time interval to get more insight.
  
• User can visualize time series based on States.
  
• User can select multiple states from the map for visualization.
  
• User can compare unemplyment rate between all the states.
• User can unselecte the selected states from the map.
  
• User can reset the visualization.
  

Virtual PC Viewer (Android Application)

Introduction

Virtual PC Viewer is a remote control system which allows the user to view and interact with one computer (“remote desktop machine”) to cellular phones (“viewer”) using wireless internet connection. A viewer is provided on the cellular phone that enables the user to see and manipulate the desktop of various remote systems such as Windows operating system. The system to be accessed must be running a Virtual PC Viewer server and it must be attached to a network.

Description

Virtual PC Viewer is a client server mobile application and is an ultra-thin client system based on a simple display protocol that is platform independent. This software allows to communicate with the remote system and can have control on the peripheral devices like key board, mouse from the cellular phones. The remote system’s administration tasks can also be performed from the cellular phone. Popular uses for this technology include remote technical support and accessing files on one's work computer to the mobile device. Here the mobile device will act as client and remote machine is called server The user of Virtual PC Viewer has to register at client machine before connecting to server. Registered user can connect to the remote desktop machine through IP address and PORT number of remote desktop machine. After successful connection user can access remote desktop machine. Android native code is used to make client-server communication and better User Interface is provided to make the application user friendly.

Application

One of the primary application of this project is being able to access, operate a computer remotely from the mobile as if remote machine is in your hand, User can access virtual files from remote machine to mobile device.

Doodle Blink (Android Mobile Game)

Introduction

Now days the ease for playing software games has increased tremendously. Due to busy schedule and nuclear families people op to play software games rather than playing manually. Especially those which are quickly accessible via computers or mobiles. This game is one among such games which is fun filled, probabilistic in nature. The Doodle Blink is a skating game, the doodle will always be blinking, and instead the user needs to jump Doodle to skip the obstacles. When game gets over will get score in score board and it display only highest score.

Description

The Project “Doodle Blink” is a test of accuracy, strength and concentration. This is developed platform HTML5 and with the support of Phonegap. Doodle Blink game is the task completion game in which the doodle (person) will be always blinking, the doodle will be on the skating board will be with the doodle till the end of the task competition, for each of the task completion there will be the restricted time, the task contains the obstacles etc. Doodle blink a contiguous game in which doodle is always blinking and Instead the user needs to jump doodle to skip the obstacles and in the first level obstacles are less and speed is also less and we have to score 1000 points then will go for second level in this obstacles are more and speed is also more and for going the third level we have score 3000 points and in this level we have collect the positive points. In score board consists of highest top 4 score will be displayed.

Know before you buy (Web Application)

It is a website suggests user the products (Laptop or its accessories) from different brands based on his specifications (cost, brand, model). This website gives the complete idea regarding the latest products in the market.

File Converter (Java Application)

PDF is global standard for capturing and reviewing rich information. It is more secure and dynamic than ever. Basically PDF files are read only , which are not editable so we need to convert this file to some other format which provides more editing options one of the format is Microsoft word.

File Converter is a Windows system program. It supports 98/2000/XP/2007/2008 platform this software is developed on Java platform using Itext jar, This allows user to convert PDF file to document format file and vice versa and also PDF to TXT format. This software has the ability to retain the original layouts of the source file in the conversion.

Application:

• Image extraction from PDF file.
  
• Concatenation of two PDF files.
  
• There is no problem in image processing.
  
• Splitting of one PDF file to two PDF files.
• Converting PDF file to Text file.

QuesZen

Introduction

In Education Centers the most required task is preparing question papers for internal, mid terms and final exams. Which need to meet in time and most secretive. Even university papers which are printed and sent to respective colleges have a chance of leak at last movement.

Description

QuesZen is developed on platform VB.net for frontend and MS Access as backend application and crystal report for generating the question papers into document format. The Software that generates the question paper by picking the questions dynamically from Question bank Database at the specified level of difficulty papers will be more secretive and it will be in time. It provides 3 levels of paper generation i.e. easy, medium and difficult levels. It allows selecting subject, level number of sections and number of questions of specified marks, it will generate different question set of each press of generate button by picking random question from the database and also allows you to change particular question from the generated paper, and provide you option save the 3 sets paper for each exam type.