Hey, I'm
DHAVAL POPAT
I graduated in 2019 from New York University with a Master's in Computer Science. At NYU, I worked on projects ranging from Cloud Computing, Backend Development to Artificial Intelligence, Data Science, Computer Vision, and Deep Learning.
I have completed my Bachelor's in Information Technology from the University of Mumbai, where my work in senior year project led to publishing a research paper in Springer.
I am enthusiastic about working on projects that require me to step outside my comfort and knowledge set, as it is important for me to continue learning new development tools and techniques.
I like being challenged and have always been dedicated to solving any roadblocks that came within my way to fulfilling a particular task.
Throughout my professional experience, I have found that being able to suggest innovative solutions for complex problems has been invaluable in my professional development.
Outside of academics, I have participated in Trivago Hackathon to build an intelligent assistant where I was responsible for implementing the text summarizer module, and also enjoyed processing the data.
I love to draw in my free time. I am also trained in Karate Budokan.
At Microsoft, my focus revolves around working with stakeholders to design and implement functionalities leading to increased adoption of Azure Stack HCI.
My work at AWS in the EC2 org gave me an opportunity to learn a variety of skills including building distributed and scalable systems. My first project soon after joining was to build an automated remediation system to reduce unsellable capacity across EC2 fleet, recovering hosts worth $100MM CAPEX. I also worked on reducing operational burden and investigation time by 50% for real-time logs for one of the critical service that stores inventory of all the EC2 hosts in the world and eliminated unnecessary infrastructure maintenance. Furthermore, I bootstrapped micro-services in new AWS regions responsible to build every EC2 host worldwide and perform all the software validations. Finally, I also lead design reviews, operational reviews, and provide guidance and mentoring to interns and new hires.
The summer 2018 internship at Zebra Technologies gave me an opportunity to work on a legacy system upgrade from desktop to a web application using ASP.NET MVC for seamlessly processing inventory orders. Throughout the internship, I worked on most phases in the Software Development Life Cycle, from Requirement Analysis, Planning to Design, Development, and Deployment. The agile development methodology was used to analyze client’s requirements and incorporate advanced features that improved forecasting accuracy. I also implemented innovative functionalities that led to about 65% increase in user productivity and designed SQL tables to efficiently support them. My work with integration of smart calendar controls and devising an intelligent carrier algorithm enhanced the shipping effectiveness.
While working on-campus at NYU Langone Health (School of Medicine), I collaborated with an interdisciplinary team to provide innovative technological solutions. I worked on the development of an interactive desktop application to facilitate cognitively impaired patients with text and voice based communication. The cranial placement of an experimental treatment tDCS device was verified by fine-tuning the Faster R-CNN model with Inception v2.
During my summer 2017 internship at Drishti, I worked on building a real-time surveillance system that detects and tracks people in deep sea water to prevent them from drowning. I was responsible to conduct research and explore various algorithms to built the application with good performance. The system had to be trained with about 30,000 data samples using cascade classifiers that resulted in achieving over 85% detection accuracy. A module was integrated to alert lifeguards by sending them the missing person’s location and tracked route.
Amazon Web Services have been used to develop this scalable conversational dialog engine that recommends restaurants to users based on their location and preferences. I have used AWS Cognito to authenticate the users and IAM roles to manage authorizations which help users to access various services. The first Lambda function utilizes Lex to generate responses and communicate with the user. Another event driven Lambda function is employed to fetch suggestions from the Yelp API and push them to SQS queue. The final Lambda funtion is triggered in a timely manner to retrieve recommendations from SQS and provide them to the user using SNS.
Event Finder is a web application that displays events happening in the local neighborhood by populating the map using D3.js based on user’s preferences. The local events data is fetched by hitting an external API using basic auth. I implemented the authentication module and created a middleware function to authorize users’ API requests. I also built an API endpoint to get information of users and restricted its access by verifying user roles with admin privileges.
The goal of this project is to apply various masks on facial images using landmark detection and mask morphing. The facial landmarks are retrieved by segmenting skin region from the prominent facial features and analyzing its frequency projections. Further, I morphed various masks using non-linear alignment of images with warping to match relevant landmark points and fit the face. The resultant masks are overlaid onto facial image using Lanczos interpolation and alpha blending giving an effect of augmented reality.
This project gave me an opportunity to explor and preprocess sensing data gathered from accelerometer, microphone, light sensor, app usage, GPS, etc. to develop sleep model, conversation classifier and activity predictor, and integrated this data with academic workload. I analyzed behavioral changes using visualizations to interpret relationship between smartphone data and student’s mental health state. Further, I utilized my knowledge of data reduction, predictive modeling, and model selection & evaluation to build a linear regression model with lasso regularization for predicting depression and stress levels.
X-Beats is a desktop application that generates an ideal music playlist for the user aiming to soothe their mood. The user's facial expressions are captured to predict the emotional state using Viola Jones algorithm and AdaBoost Training. I also implemented a module to track all processes running on the system by utilizing WMIC commands. This enhances the playlist by including music that maximizes productivity pertaining to the application being used.
This desktop application anticipates price of an airline ticket by employing Linear Regression. It also predicts ticket availability on a particular date in future by implementing Naïve Bayes classifier.
The purpose of this project was to spread awareness about reducing, reusing, and recycling. This website facilitated users to upload and share images of trash littered on streets with local authorities. Google Maps API were utilized to provide assistance in locating nearby reuse and recycling centers.