About

I am working as a Machine Learning Engineer at Amazon Personalization where I primarily work on Recommendations.

Education

I have a Master’s degree in Computer Science from University of Georgia (Graduated in 2019) where my primary focus area was AI and Machine Learning. I completed my Bachelors degree from Nirma university, India in 2013.

Academic experience

All of my Data science/Machine learning projects can be viewed at https://ankit-vaghela30.github.io/machine-learning/

Work experience

Machine Learning Engineer at Amazon, NYC, NY

Ran ranking experiments on Amazon’s desktop and mobile homepage to rank content which resulted in +15 MM GCCP (Growth adjusted Composite Contribution Profit) for desktop and +100 MM GCCP for mobile homepage.
Developed hypotheses based on previous experiments to improve model performance and added feature generators and new objectives to existing featurizer flow. As a result Improved model performance for Non-Prime customers and not logged in customers.
Collaborated with the team and published a research paper in an internal AMLC conference based on above experiment findings.
Lead SDE from my team for Prime day 2023 where I led the efforts to scale up services, switch model objectives from GCCP to total sales and the drafting process of validating the switch and model’s exploration.
Responsible for heading efforts to migrate our service from Intel hosts to AWS’s new graviton hosts which resulted in saving $100 per month and lesser carbon footprint (more energy efficient)
Responsible for switching our service’s loadbalancers from network level Netscaler to internal application level tardigrade loadbalancers to decrease costs. Also, switched service authentication to internal Cloudauth authentication to make service more secure (HTTPS).

Data Scientist at ADP, Alpharetta, GA

I have been working at ADP as a Data Scientist. Responsibilities:

ML privacy and security: I created an API for automatically detecting sensitive PIIs from pdfs and mask them. Used NER model Spacy for identifying PIIs from the pdfs.
AI for QA: Worked on a model to select which tests to run based on git changelog which reduced build time by 30%.
Responsible for creating BDD based end to end testing framework for Machine Learning pipeline.
Code quality lead: Responsible for code reviews and making sure code standards are maintained in the repo.
Worked with fellow Data scientists to help them integrate and deploy Machine Learning models in the pipeline.
Responsible for designing a Model drift and data shift application which can automatically trigger model retraining.
Worked on “Handwritten text recognition” where I used Deep learning models to identify characters and text from handwritten forms.

Software Engineer at NCR, Atlanta, GA

I worked on Fuel Kiosk software for one of the NCR’s self serving kiosks product for gas stations: Optic. I created a module in C++ to fetch Customer loyalty information using web service and pass it on using Google Protobuf messages. I have created BDD automated tests in Python and modified existing Javascript based application to display the loyalty information. I also modified UI of the application using HTML and CSS.

Research Assistant at Schmitz lab, UGA, Athens, GA

I worked as a Research assistant in the capacity of MEAN(MongoDB, Express.js, Angular.js, Node.js) stack developer at Schmitz lab. I have created a web based platform which is being used by UGA’s undergraduate biology course students to learn about genetics and inheritance through simulated experiments.

My work can be viewed by going to the lab website and go to Mendelpede application. link: https://heredity.genetics.uga.edu/

Software engineer intern at Verifacto Inc., Atlanta, GA

Verifacto Inc. provides a middleware between Vehicle insurance providers and their customers (insurers). I worked here as a Software engineer intern. Data collection and integration is the main function of the company. Multiple Vehicle insurance providers like Statefarm, All state etc. sends insurance information via various sources like email, Postal mail and even sometimes photo clicked from phone. My task was to collect all these data and integrate it in one place. The pipeline worked in a way such that all the documents are converted to physical copy which is then scanned and inserted into local file system. I extended this pipeline by running OCR(Optical Character Recognition) on it, then storing the text file on Amazon S3 server, then running customized regex for each vehicle insurance providers to collect the data and insert it in MySQL database. This data was used to provide a web based interface to both service providers and customers.

Technologies used: Python, MySQL, BeautifulSoup, Data integration

Associate of technology at Cognizant ltd., India

KPN, Netherland

KPN is a telecom giant based in Netherland. I worked on order fulfillment project in Salesforce framework. I worked in the capacity of Salesforce developer and created Visualforce pages, Apex triggers, Apex classes etc. I also worked with the product mowner to develop encryption/decryption feature.

Technologies used: Salesforce platform, Vlocity

Senior systems engineer at Infosys ltd., India

After completing my Bachelors degree in Instrumentation and Control engineering, I grew interest in programming so I joined Infosys ltd as a Software engineer. I worked for multiple clients on multiple projects which are described below:

First Data, USA

I worked on 1099 Tax reporting application which was a inhouse reporting application used by First Data employees. I created a functionality to mask and unmask sensitive customer information (i.e. Tax Identification Number). Additionally, I developed a login based privilege management to allow appropriate users to view sensitive information. User behaviour like who has unmasked sensitive information at what time, was also logged on servers using a logging tool splunk. I also supported security testing of this application and rectified some of the vulnerabilities like URL injection.

I also supported direct client requests which mostly asked to create database scripts for data insertion/updation. I observed that this task can be automated by creating a Java tool to read excel sheet, manipulate the data and output a SQL script. I created the tool which resulted in saving 90% of manual labour by providing the interface directly to First Data executives.

Technologies used: Java, J2EE, Struts, Linux

American Express, USA

I worked on a Big Data project to create a system which showed credit card usage report in Hbase database. I created a data pipeline to filter customer data using pig, hive and map reduce algorithms.

Technologies used: MapReduce, Java, Pig, Hive, HBase,

Telenet, Belgium

Telenet is a telecom giant in Belgium. My first project for Telenet was a Order fulfillment application where I worked on workflows, Apex triggers and Apex classes. It was entirely built on Salesforce platform where we called few Java APIs for billing process.

I later joined the Billing team and led team of four fellow colleagues to create a Java web service. For this project, I was involved in each stage of SDLC(Software Development Life Cycle). I interacted directly with Product manager to understand all requirements and provided estimates for the project. I also got an opportunity to perform duties of a Scrum master. I along with my team created this web service for the billing application which provided access to variety of data. I used Oracle service bus which is a SOA solution of Oracle.

Technologies used: Java, SOAP, Oracle service bus, Salesforce