KUMAR SAHARSH
I am Computer Science Enthusiast who likes to work on real-life projects. I like to explore and work on different domains of Computer Science such as development, machine learning, Competitive Programming. Right now, I'm more inclined towards Machine learning and do some more cool projects. I'm the type Of the guy who's always delighted (kinda habit, can't help it). I'm the big foodie who lives to eat. I aim to build a personal AI like JARVIS from IRON MAN series who'll take over the world for me xd.
Internships
Siemens Technology Pvt Ltd Bangalore
I worked as Summer Research Intern in my sophomore year. The work was to generate follow-up questions to extract information from experinced personnal. I experimented with open domain follow-up question generation and acieved closed domain question generation. The idea is to choose a domain and use a article to generate question about common attributes of the domain and use those questions to extract information about some other product.
Major Projects
Open Domain Question Answering System
An Open Answering system aims at returning an answer in response to the user's question using Natural language processing from Wikipedia.
- Document Retriever and Document Reader work simultaneously to generate a response.
- Document Retriever gives ten most relevant articles from Wikipedia.
- Document Reader finds the most relevant answer span from the list of articles using Attention Model.
Smile Detection
A End-to-End project which can detect if the person in the upload image is similing or not.
- Build Using Keras.
- Dataset used : The MPLab GENKI Database
- 4 ConvNets and Max pool layers are used one after another followed by a Fully connected layer.
Emojify
Emoji prediction from given a sentence using LSTMs.Completed as part of Deep Learning course. I wrote the whole code from scratch.
- Predicts upto 5 different emojis.
- Fairly Decent Accuracy of 87.5% on dataset provided in the course.
English to French Translation
English to French machine translation using Sequence to Sequence model.
- Character level Seq2Seq model Implemented
- Word level Seq2Seq model also Implemented which observed better prediction than the former.
Sentiment Analysis Using Naive Bayes
Used Naive Bayes to do Moive Review Classfication and Tweet Classfication.
- Movie Review Classfication Achieved accuracy of 71.2%.
- Tweet Classfication has a heigher accuracy of 78%.
- Both of them are binary classfication, that is positive and negative.
HomeMade Food App
A food ordering and delivery app like Zomato
- Made different layouts such as order history layout,mass order, consumer dashboard.
- Used Cloud Firestore as database
Music WebApp
A music listening app.
- Used Django and sqlite as backend. Features to mark songs as favorite.
- Still in progress.
CUBETISTIC-CUBETHON
A game made using unity as part of learning game development. A cube runner game with the aim to avoid obstacles and reach the finish line.
- 4 Levels with increasing difficulty.
- Dynamic spped , increase as level progresses.
- Level Restart in case of collision with obstacles.
- Work on Windows OS.
Experience
MACHINE LEARNING
As of Aug'19 its more than a year that I'm working on ML. I've done various projects ranging from regression models to Neural Networks. My main work is in Natural Langauge Processing and Neutral Machine Translation. I've two major projects and a internship in the same.
WEB DEVELOPMENT
This website was made by Bootstrap. Currently, I'm learning Django. I made a website from scratch as part of the learning phase(link mentioned below). I'm making a music app right now as part of Course Project using Django and MySQL as backend.
COMPETITIVE PROGRAMMING
I know the use of Data structues, Maths for Competitive Coding ,Bit Manipulation.In Data structures I'm well acquainted with use of Stacks, Queues, Linked lists, BST, trees and Graphs.
ANDROID DEVELOPMENT
I started with making static layouts in my freshmen year, followed by the use of Firebase and Firestore database and permanent data storage in my sophomore year. I've made a Zomato like app as part of a group course project using Firestore and a note making app .
ENTREPRENEURSHIP
As much as enjoy doing techincal work ,I like doing Non-technical work too.Started from the beginning of college-life, I gained much experience in marketing and Entrepreneurship.I did ISTE Marketing SMP in Summer 2018 for far more depth in marketing world. In my fifth sem I'm doing a course on entrepreneurship as part of college syllabus.
GAME DEVELOPMENT
Got introduced to UNITY as a part of tech-fest project in NITK. Link to My First Game. I found Unity to be really amazing ,with a lot of customization and features, its the most interesting thing to do in freetime. I'm planning to make Virtual Reality games in Future.
Education
NATIONAL INSTITUTE OF TECHNOLOGY SURATHKAL KARNATAKA
CGPA: 7.61(Aug 2019)
VIDYAMANDIR CLASSES PATNA
Skills
Interests
I am the kinda of guy who just can sit idle and do nothing all day, I need the kick to get my life running. That's actually one my interest , keeping myself busy with something always.
I really like doing real life projects which affects lives of people and makes difference. Playing high-resolution games is something I really likeI am the kind of person who "LIVES TO EAT".
Achievements
- Volunterred in Smart India Hackathon at NITK organised by MHRD.
- Volunterred at Women in Technology Origanised in NITK to encourage women participation in different domains of computer Science.
- Part of Hindi and Sanskrit Sabha in the freshmen year.
- 3 Stars on Codechef.
Certifications
Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization
Issue Date : August 2019
Credential URL
