I am an experienced Machine Learning Engineer, Scientist and Software Engineer skilled in building models for Natural Language Understanding, Natural Language Processing, Information Retrieval - Recommendation Systems, Search, Ranking and Computer Vision. Expertise in Generative AI, Deep Learning Architectures, Classical Machine Learning, MLOps, Scalable System Design, Distributed Systems and ML System Engineering. Passionate about leveraging technology to drive innovation and deliver high-quality solutions. Currently working at Bed Bath & Beyond (Beyond, Inc) as Machine Learning Engineer and Scientist within the personalization, discovery and recommendations team. Let's connect to explore how I can contribute my skills and experience to make a meaningful impact.
Technical Skills
Artificial Intelligence and Machine Learning
Research and Projects
This thesis is divided into two parts:Part I: Analysis of Fruits, Vegetables, Cheese and Fish based on Image Processing using Computer Vision and Deep Learning: A Review. It consists of a comprehensive review of image processing, computer vision and deep learning techniques applied to carry out analysis of fruits, vegetables, cheese and fish.This part also serves as a literature review for Part II.Part II: GuavaNet: A deep neural network architecture for automatic sensory evaluation to predict degree of acceptability for Guava by a consumer. This part introduces to an end-to-end deep neural network architecture that can predict the degree of acceptability by the consumer for a guava based on sensory evaluation.
Lending is one of the key business areas in the banking industry, credit cards as of late have seen huge success over the course of years. In pursuit to increase their market share,banks often issue credit cards to ineligible customers without adequate background checks. Also, many customers used their credit card beyond their repayment capabilities leading to high debt accumulation. Identifying the risky and non-risky customers is the biggest challenge for banks. So, the problem we are trying to analyze is how to identify the risky and non-risky customers, helping the bank to decide if a customer has the potential to repay the used credit of the bank.
This information engine is able to tell its users what is happening around in RIT by providing them with the top terms from the collected documents. The machine learning model is generated using Simple K-Means document clustering utilizing the scikit-learn library in python. The model is deployed using RPC (Remote Procedure call) as a distributed system for the purpose of making a client-server architecture and is utilized by a web-application based on Node.js/Express.js. Further, the search engine utilizes OKAPI BM25 algorithm to search through words in the clustered documents.
As a part of mini project for client design, the requirements were to manipulate DOM (Document Object Model) and do things with form elements on fly using only JavaScript and related technologies in their Vanilla form. For this project we were not allowed to use innerHTML. This mini app finds your Body Mass Index consuming a third party API. It takes in your body vitals posts it to a service via AJAX which makes cURL request to the third party API and returns a response in JSON which includes BMI calculation results. The app utilizes SVG for graphics, PHP for service and the front-end is based on bootstrap without JQuery.
This paper explores a very well known technique for image classification and recognition that is Bag of Visual Words. The process involves feature extraction using Canny Edge Detec- tor and Scale-Invariant Feature Transform (SIFT), codebooks construction using generative model like K-Means and Vector quantization. Finally, classification is done using Support Vec- tor Machines (SVM) using chi-squared kernel. After applying 10 cross validation the accuracy comes to be around approx 70%.
As a part of requirement for Knowledge Representation Technologies project. We were required to utilize NoSQL graph based database Neo4j and develop a web/mobile application. Path Finder is an application to find shortest path between a source room and a destination room in the B.Thomas Golisano college of computing and information sciences 2nd floor (IST department). Uses Dijkstra algorithm to find all shortest paths. Technology stack used is Neo4J graph database for the back-end and the front-end utilizes Node.js, Express.js, JQuery and HTML Canvas.
As a part of requirement for Client Design and Development, developed web-presence for the IST (Information Sciences & Technology) department of Rochester Institute of Technology. Re-usability and usage of JQuery was the main requirement for this project. Consumed existing API using JQuery AJAX and generated the content dynamically for the website.
Forecasting the exchange rates is both a challenging and important task for the modern traders, people working in the financial markets and general population across the globe. In this paper we will be utilizing the time series concepts to do an analysis and predict the daily exchange rates of the Indian Rupee (INR) against the United States Dollar (USD). This paper will investigate and compare different forecasting techniques like ARIMA, Holt-Winters simple exponential smoothing and Neural networks. Further, utilizing the above techniques investigate the behavior of daily exchange rates of the Indian Rupee (INR) against the United States Dollar. (Daily exchange rates from 19th November 2007 to 18th December 2017 were used for the analysis).
Twitter Application is a Node.js/Express.js based web application. As a part of curriculum for Knowledge Representation Technologies, task was to take exceptionally large data-set store it in MongoDB and do CRUD (create read update delete) operation on them through a client. Used EJS (embedded JavaScript) to render views, on the back-end Express.Js framework in vanilla form over the top of Node.Js. Frontend was designed using Twitter Bootstrap along with a server side implementation of JQuery Data-tables and jQuery.
