Pooja Ramapurath

Successful at using programming technologies such as Java, MySQL, and Python to develop code. Disciplined and reliable professional with excellent critical thinking skills and problem-solving abilities.

• TechSoc Coordinator (2021 - 2022)
• Tensors Associate Coordinator (2020-2021)
• Tensors Consultant ( 2019-2020)

Sentiment Analysis of Twitter Data: Implemented a binary text classifier (tfkeras LSTM layers) to classify the sentiment behind Twitter texts.Achieved 93.6% accuracy on validation of the model with test tweets with similar/same recognized subjects. Incorporated analyzed sentiment of Twitter users to predict future price movements of respective subjects.  

Algorithmic Trading Automated Script :Designed a deep learning model using custom time series indicators as input; achieved average accuracy of 85%., Modeled custom output for effective ANN training; developed strategy to use predicted outputs for profits. Developed script to adhere to strategies, place appropriately leveraged positions & corresponding TP/SL on Bybit Achieved net profit of +27% (incl. fees deduction) in 3 weeks with 26 profitable trades out of a total 30 initiated.

 Optical Character Recognition: Leveraged OpenCV to correct text skew instituted statistical methods to segment paragraphs into characters.Trained CNN architectures on 70,000 strong MNIST dataset of handwritten numbers, using TensorFlow 2.0. Achieved maximum accuracy of 95% with a 6-layered CNN, further to 97% incorporating Dropout Regularization

Credit Card Fraud Detection :Built model to predict credit fraud of ~50K customers on a dataset of American Express (0.3Mn transactions). Used oversampling, SMOTE for dealing class imbalance, & deployed gradient boosting algorithm XG Boost. Achieved a distinctive F1 score of 0.70 in identifying fraudulent transactions using a 30x1 feature vector. 

Heart disease risk prediction :Created a Python script with Feature Importance and EDA to predict cardiac issues using a heart disease dataset., Implemented algorithms like Logistic Regression, K-NN, SVM and accomplished 90% accuracy with RF. 

Customer segmentation :Market Basket analysis is performed to assess the customers' choices so as to define and meet their needs, Performed hierarchal clustering on Customer Reviews, derived optimal clusters Dendogram Threshold Approach. Used techniques like Text Mining, Topic Modeling using Unsupervised ML algorithms like K Means

• Cleared JEE Advanced in the first attempt and 98.66 percentile in IIT- JEE Main from over 1.5 million applicants.
• Received cash prize from honorable District collector for being in the top 5% students with high marks in class X, as a part of Sahodaya.
• Got all Kerala 3rd rank in the All Kerala Sahodaya Scholastic Achievement test conducted in 2013

• Trained in music for a year and painting for over 6 years.  

• Shaastra 2019  volunteer, organized and managed the games, and leisure activities for 100 plus students