I am a Electronics and Communication student at Delhi Technological University (formerly DCE). I enjoy solving software problems with Machine/Deep Learning and Artificial Intelligence. Currently, my work is focused on effective computing and how to detect micro-expressions. Actively looking for SDE full time oppurtunities.
Utkarsh Bisaria
JCB - 126
Delhi Technological University, Delhi India
+91-9457996495
2utkarsh@gmail.com
B.Tech in Electronics & Communication (2016-2020)• GPA : 8.60
Foundation in fundamental Electronics, Computer Science Engineering concepts. The course includes subjects of Data Structures and Algorithms, Operating Systems, Database Management Systems, Digital Image Processing and Machine Learning.
Senior Secondary School in Science • Percentage : 95.4%
PCM with Computer Science in 12th grade. Highest overall percentage in a batch of 200 with top scores in Chemistry, English and Maths. All India Rank 6653 in JEE Mains with 12 lacs+ participants.
Software Developer Intern• May 2019 - July 2019
NamedEntityRecognition - Implemented named entity recognition using spaCy python library for the purpose of automatically summarizing documents such as resumes,billing documents of assets,for improving customer support.
I am a self learnt web developer and Machine/Deep Learning enthusiast with a keen interest in solving software problems. Miles to go before I sleep .........
Developed English to French human language translator using a Recurrent Neural Network based on Long Short Term Memory Model. The network was build using encoder-decoder architecture and trained on Tensorflow Python Framework.
Natural Language Processing , Machine LearningDeveloped a Captcha Breaking model based on Deep Convolutional Neural Network which tries to identfy the letters in captcha and hence break it. The network was build using convolutional layers, max pooling layers and fully connected layers in Keras Python Framework. Preprocessing of captcha images was done using OpenCv.
OpenCV, Convolutional Neural NetworkBuilt a movie recommender model using collaborative and content based filters. For content based approach, top 250 rated movies from IMDB were used and similarity matrix based on cosine similarity. MovieLens dataset was used for collaborative filter approach and KNN model was built to make recommendations for top 10 similar movies.
Natural Language Processing , CountVectorizer , K-Nearest NeighboursUsed Twitter’s tweepy api to fetch live tweets to train a classifier using Support Vector Machine, Naive Bayes, Stochastic Gradient Descent algorithms to classify a tweet into positive or negative sentiment and graphed live tweet sentiment analysis. This was done using Scikit-Learn classifiers in Python.
Natural Language Processing , tweepyBuilt a future price movement predictor for four cryptocurrencies using Recurrent Neural Network based on CuDNN Long Short Term Memory Model. Created sequences of 60minutes of data which were used in RNN model to predict price movement of 3minutes in future for each currency. Implemented using Tensorflow Python Framework and visualized the in-sample and out-sample accuracy using TensorBoard.
Recurrent Neural NetworkBuilt a weather prediction supervised model using Random Forest algorithm to predict maximum temperature in a city based on one year of historical maximum temperatures (regression approach) and average temperature for last two days.
Random Forest , Machine LearningAnalyzed tabular and geospatial data from Indian Space Research Organization to develop different metrics for ranking individual villages of the country. Used KMeans clustering method to discover villages which share similar set of problems. Developed an android application to allow regular citizens of the country to adopt villages and work on their improvement. The project was made for the Smart India Hackathon 2018 Grand Finale.
Data Analysis , Machine LearningThe website gives information regarding best camping sites in the world. Website allows to create account using and bookmark the favourite places. Stack for website used is NodeJs, MongoDB, HTML, CSS, Bootstrap, Javascript,RESTful APIs.
Full Stack Web DevelopmentWebsite and android application was made to change the TV channel using computer vision, youtube API. Android application was used to detect the short and long blink of the user and decode the morse code so that the specially abled people can change the TV channel with blink of eye.The project won first prize in hackathon conducted by DISHTV.
Data Analysis , WebappFeel free to reach out to me at 2utkarsh_bt2k16@dtu.ac.in, 2utkarsh@gmail.com
Until then Adios Amigos