Harsh Tamkiya
About me
I am Harsh Tamkiya. I am a final year student, studying Computer Science from Indore. I have keen interest in designing algorithms and solve data related problems. I love spending time on Hackerrank, Kaggle.
I am pursuing my undergraduate degree in Computer Science from Shri Vaishnav Vidyapeeth Vishwavidyalaya. I have learned Data Structures and Algorithms, Algorithm Design, Data Science, Artificial Intelligence, Deep Learning, and Latest Cloud Technologies like AWS SageMaker.
In my 4 years of engineering, I have actively participated in various Hackathons, Research Seminars, etc. Being an active participant of Google Developer Group, Facebook Developer Group in my city, I have attended various coding workshops on the latest technologies. I have published my first Research Paper titled as "Candidate Background Verification Using Machine Learning and Fuzzy Matching" on June, 2020.
I building new stuff. You can visit my GitHub profile for all my project works. Connect with me on Linkedin and Twitter. Feel free to ask me anything about Data Science, Careers, Python projects and Competitive coding.
Publications & Research
Candidate Background Verification Using Machine Learning and
Fuzzy Matching
Himanshu Suman, Harsh Tamkiya, Anurag Singh Kushwah
International Journal of Emerging Technologies and Innovative
Research (JETIR).
Internship Experience
Software Engineer Intern
During my internship at Techrific Technologies, I worked on collecting, studying and interpreting large datasets; prepared reports; performed accurate and successful data management tasks. Built fuzzy matching algorithm to identify, validate and classify genuine candidate profiles. Collaborated with creative teams to develop an automated resume verification system. Collaborated with research and development teams to develop new forecasting models using ensemble learning and neural networks and increased the performance of models.
Machine Learning Intern
During my time at WittyFeed, I worked on Recommendation System for a video based application. Did research on types of recommender systems and choose collaborative filtering for designing user-item based recommendations. Used Alternating Least Squares (ALS) algorithm as it performed really good for implicit dataset. Implemented ALS model in PySpark framework.
Selected Projects
Built an RNN performing sentiment analysis on movie reviews complete with publicly accessible API and a simple web page which interacts with the deployed endpoint. Deployed this application on AWS. Used AWS Sagemaker, AWS Lambda and AWS API gateway for building this project.
In this project, I took Bangalore, India house prices dataset from Kaggle. I have Analysed data carefully and designed different python functions to clean data with keeping in mind every outlier. Used Grid search CV to get the best performing model for the dataset. Created a web application on Flask for getting predictions.
I was curious about how insights are collected from data, so I took the uber rides dataset of April-2014 of the city Manhattan, NY. Visualized the Time Series dataset using different python libraries. From the Heatmap, Bargraph and Lineplot found some interesting insights. 'People took uber rides more on weekdays than weekends.' '3 P.M. to 9 P.M. was the prime time for uber rides.' 'Friday nights were the most busiest nights of the month.'
As a part of Kaggle competition Zero to GANs - Human Protein Classification, I have worked on multilabel image classification problem, where each protein image can belong to several classes. I used Resnet-34 architecture with One cycle policy. Also used data augmentation, data normalization which significantly reduced training loss.