Parth S.

ABOUT ME

I have been into programming since teenage. Always ready to explore new technologies. Willing to engage in some serious work. I have been working in Machine Learning, Deep Learning, and Data Science from past two years and one of the contributors in Scikit-Learn.


My past projects technical domains:

Linear regression, Logistic regression, PCA, Backpropagation, Autoencoders, Generative adversarial network, Dropout regularization, K-means, KNN, SVM, Object detection, Classification, Localization, Segmentation, YOLO, OpenCV.

* Natural Language Processing (NLP)
-- intent, question answering, e-commerce search, conversations. DialogFlow, wit.ai

* Time series analysis, prediction -- medical signals, stock/currency markets.
* Cloud (AWS, Azure, GCE) based network training and deployment.

Hands-on expertise with a range of deep learning tools:
* Tensorflow, Keras and PyTorch expert using CUDA backend.

Current Work:
• Building predictive models and data analytic tools on several oncology clinical datasets to predict required surgeries and therapies.
• Automating clinical trial eligibility process for oncology protocols and inferring for their eligibility impacting 50-60% time reduction as compared to manual inference.
• Building handwritten text OCR on several handwritten and printed text for various leading financial, healthcare and e-commerce firms
Building predictive models



Highlights:
• Image Recognition: Built models with TensorFlow on top of Inception-v3 model and CNN algorithm to extract attributes of images from few fashion retailer catalogs. Automated binning of images from fashion websites to be used for training the model.
• Color-Detection: Detect the most dominant color from images using image segmentation and machine learning techniques from fashion catalog images.


Proficient in building/training cutting-edge, novel neural architectures across various verticals. Familiar with the latest research involving DNN, CNN, RNN / LSTM, seq2seq, attention models.

SKILLS