DHEEPTHA BADRINARAYANAN Dallas, Texas
ac8ihm@r.postjobfree.com +1-469-***-****
LinkedIn github
EXPERIENCE
Intern, Machine Learning: Upstream Tech Aug 2018 –Present
Technologies: Keras, Tensor board, CNN, Rasters, Git version control
Worked on removing speckle noise in image tiles from obtained Synthetic Aperture Radars using Convolutional Neural Networks (CNN)
Conducted proof of concept to identify an appropriate network architecture to train a deep learning model without having to use clean targets. Obtained a fairly clear version of the noisy tiles with PSNR value nearly 32dB.
A short write-up on my work: https://medium.com/upstream/denoising-sentinel-1-radar-images-5f764faffb3e
Currently working on sentinel-1 and sentinel-2 data fusion to obtain better Landcover mappings. Using Convolutional Neural Networks to perform mapping between Sentinel-1 and Normalized Vegetation Index data.
Intern, Machine Learning: Verint Systems, WA Jun 2018 - Aug 2018
Technologies: Python, Keras, scikit-learn, NLTK, GloVe, Linux, Pool, Flask Rest-API, Docker
Trained a multi-class, multi-label, powerful LSTM classifier to perform Sentiment classification and POS tagging based off a virtual agent’s chat corpus data. Also trained phrasal Glove embeddings to be used as input features to the model
Improved the model’s performance using Active Learning. Delivered the functionality as a REST API service.
Developed a scalable, distributed query strategy technique to sample from nearly 20M unlabelled chat data records and improved run time by 45% using parallel processing
Research Intern: Centre for Development of Advanced Computing (C-DAC), India Dec 2016 - Apr 2017
Technologies: R, e1071, JSP, Java Servlets, PostgreSQL, Android Studio
Designed and implemented Risk Based Authentication system (Anomaly Detection) which improves online security using One-class Support Vector Machine, Naïve-Bayes model.
The authentication token was implemented as an android application
The project was presented in the 14th IEEE International Conference on Advanced Trusted Computing, San Francisco, USA. Had rave positive reviews at the conference
TECHNICAL SKILLS
Machine learning algorithms: Bayesian modelling, Random forests, Logistic regression, Markov Random Fields, Conditional Random Fields, Gaussian Mixture Models, Decision trees, LSTM, CNN, Gated Recurrent Units, Word2Vec
Python Libraries: Numpy, Pandas, Scikit-learn, Keras, Tensorflow, Matplotlib, Tensorboard
Miscellaneous: R, Android Studio, Java Servlets, REST API, MATLAB, Git
ACADEMICS
University of Texas, Dallas Aug 2017 – May 2019 (Expected)
Masters in Science, Computer Science
Relevant Coursework: Probabilistic Graphical Models, Natural Language Processing, Convolutional Neural Networks, Machine Learning, GPU programming, Statistics for Data Science, Algorithms and Data Structures
SASTRA University, India Jul 2013 – May 2017
B. Tech, Information and Communication Technology GPA: 8.46/10
PROJECTS
Churn Modelling using ANN: Trained an Artificial Neural Network using a probabilistic approach to predict if a bank customer will leave the bank or stay. Implemented using the TensorFlow and scikit-learn packages in python and achieved an accuracy of 87%. Tuned hyperparameters using grid search
Structure Learning: Learnt a discriminative Bayesian Network to predict poisonous mushrooms using the Chow-Liu Bayesian structure learning algorithm. Implemented in MATLAB.
Shape recognition using CRF: Trained a Conditional Random Field Model to perform a computer vision task of shape recognition. Implemented using python
PUBLICATIONS
“A novel edge-based embedding in medical images based on unique key generated using Sudoku puzzle design”. http://link.springer.com/article/10.1186/s40064-016-3356-1
Stock market Trend Analysis: “Application of statistical methods to perform trend analysis on stock markets”. Published in the IEEE conference, INDIACom2017
http://www.bvicam.ac.in/news/INDIACom%202017%20Proceedings/Main/papers/876.pdf