About
Researcher
- City: Surat, India
- Degree: Bachelor of Engineering
- Email: 9arshit@gmail.com
My name is Harshit Agarwal and I have completed my degree in Bachelors of Engineering in Computer Engineering in 2020. I am currently working as a Data scientist at UnMazer.AI. My extensive reading over the course of my engineering helped me identify my core interests in working on Operating Systems, Artificial Intelligence and Computer Science Theory in the long run. My undergraduate studies from Gujarat Technological University exposed me to an array of courses from Theory to Programming Languages, Algorithms to Machine Learning & Artificial Intelligence. Such broad coverage of the program, along with extensive laboratory assignments have helped me develop strong foundation for my research focused career in future.
Skills
Languages
- Python
Cloud
- AWS Athena
- AWS BeanStalk
- AWS EC2
- AWS SageMaker
- AWS S3 Storage
Tools and Frameworks
- Tensorflow Library
- Keras Library
- Matplot Library
- OpenCV Library
- Flask
Databases
- MySQL
- Oracle SQL
Area of Interests
Operating Systems
Financial Engineering
Artificial Engineering
Machine Learning
Deep Learning
Natural Language Processing
Data Analysis
Education
Bachelor of Engineering
2016 - 2020
Gujarat Technological University, Ahmedabad
Computer Engineering
CGPA: 8.75/10
XII - Science Stream
May, 2016
Lancers Army School
Gujarat Secondary Education Board
Percentage: 92%
Achievements & Leadership
Winner of Flash Mob
2017
Surat
State Level Dance Competition
Volunteer
May, 2016
Disha NGO, Surat, India
Volunteer Coordinator for the NGO engaged in helping Autistic and other special children
Professional Experience
Data Scientist
June, 2021 - Present
UnMazer.AI, Banglore, India
- Data Cleaning and Extraction.
- Developing footfall analysis module for tracking any locations visit information.
- Data analysis on geolocation data for providing better information and knowledge to client for business intelligence.
- Working on Synthetic Data Generation for user specific GPS location.
Team Leader – Learning Management System
March, 2021 – June, 2021
Ignitus, Pittsburgh, PA
- Developed e-learning contents and software modules backing the Ignitus Learning Management System to be offered to the University of Michigan students.
Undergraduate Research Assistant
Jan, 2020 – May, 2020
Sarvajanik College of Engineering and Technology, Gujarat, India
- Assisting in the research project related to Music Analysis and Generation using GAN under Prof. (Dr.) Keyur Rana.
Data Science Intern
May 2019 – June 2019
Krupa Diam., Maharashtra, India
- Team member for building predictive model for company’s sales, inventories and budgetary.
- This system helps management to study and analyse their sales trend prediction and based on that their inventory can be managed.
Publication
Analysis and Prediction of Stock Market Trends Using Deep Learning
The paper proposes a progressive conclusion on the application of recurrent neural network in stock price forecasting. We have also used random forest classifier to the factor in the sudden fluctuations in stock prices which are the derivatives of any abnormal events. Machine learning and deep learning strategies are used by many quantitative hedge funds to increase their profits. A good and effective prediction system can help investors and traders to get a glimpse of the future direction of the stock. The non-linearity and chaotic nature of the prices can be combat using recurrent neural networks which are effective in tracing relationships between data and using it to predict new data.
Analysis of Process Scheduler Using Neural Network in Operating System
Process scheduling plays a vital role in multitasking for any operating system. There are many factors involved during process scheduling like priorities, free memory, user demand and processor which if not handled properly can be very complex and time consuming. Neural network has adaptive nature which can be used handle the complex part easily. The main aim of this paper is to review different types of scheduling algorithms working on the principle of neural network and offer constructive criticism to improve their efficiency.
A Neural Network Based Approach for Operating System
The operating system is the central element of a computing device. It is the base level on which all the applications run. It allows the user to interact with the hardware with the help of the user interface. Creating a more efficient and capable software means less load on the hardware. Evolving nature of the neural network will help the operating system to learn about the user and will help in creating a better experience for the user. In this paper, we propose the integration of the neural network system at the kernel level of the operating system. Further, we show that the proposed scheme is more efficient and advanced than the current conventional system.
Sentimental Analysis of News Headlines for Stock Market
Predicting the stock market is like one of the most common things one does. Predicting exact value is like a futile attempt but one thing is sure we can know the trend or so called the direction in which the price of the stock will move. The stock market is driven by a lot of factors and the majority of them are in the form of news articles. The impact of news on stocks is quite significant. The direction of stock price will be dictated by the sentiments of the market participants. The sentimental analysis is no new topic people have been doing it for quite a long time by implementing several models. In our research paper, we are comparing the results from different models under the same circumstances and concluding which one of the compared models is better based on their accuracy. The methods that we have used for comparison are K-Mean clustering, Naïve Bayes, and Support Vector Machine. In our experimental study for sentimental analysis for news headlines, we found that Support Vector Machine and Naïve Bayes have better accuracy than K-Means clustering.