I'm Harshit Kumar (हर्षित कुमार).
I enjoy learning regardless of subject; sometimes I find myself studying microeconomics but my primary interest lies in machine learning and data science. I love finding patterns in data and solving problems. I've worked on various machine learning problems in computer vision and natural language processing. In my free time, I enjoy going through Q&A on the StackExchange network and playing sudoku.
Currently, I'm in senior year pursuing B.Tech (Bachelor of Technology) in Computer Science and Engineering (CSE) at Ambedkar Institute of Advanced Communication Technologies and Research (AIACTR), GGSIPU, Delhi.
If you like poetry, check out my favourite collection.
- Programming/Markup languages:
- Machine Learning:
PyTorch, Keras, OpenCV, sklearn, pandas, numpy, matplotlib
- Tools and Frameworks:
- AI (Computer Vision) Intern Jun 2019 - Present
Vehant Technologies Pvt. Ltd.
- Working on one of the remarkable applications of Computer Vision - essentially Semantic and Instance segmentation in Video Analytics.
- Worked on Pedestrian Attribute Recognition problem. Technologies used: PyTorch, TensorFlow, OpenCV
- AI (NLP) Intern Jun 2018 - Aug 2018
Arbunize Digital Pvt. Ltd.
Applied Natural Language Processing algorithms on resumes for automatic information extraction.
- Implemented Satz sentence boundary detection system using decision trees and neural networks.
- Worked on job title prediction based on skills. Developed different models including random forest, support vector machines, K nearest neighbors.
- Developed a multi-output neural network based MBTI personality prediction model based on an applicant's writing style.
- Implemented algorithms for segmentation of the resume into different sections such as Education, Skills, Experience etc.
Technologies used: Keras, scikit-learn
- Machine Learning InternMay 2018 - Jun 2018
Developed a traffic sign detection and recognition system using Convolutional Neural Networks for electric vehicles.
Technologies used: Python, Keras
- ImageCaptioner: Image captioning using Encoder-Decoder Developed image captioning application based on Neural Image Caption model utilizing encoder-decoder architecture, using pretrained CNN as encoder and LSTM as decoder.
- TweetSense: Real-time social media sentiment analysis : Developed an Android app that analyzes citizen's tweets (using Twitter streaming API) and intelligently provide real-time feedback using sentiment analysis (natural language processing).
- DengueApp: Location based dengue prediction : Developed an Android app that gives real-time location-based dengue risk index to the user using machine learning techniques.
Check PROJECTS page for more details.