Shrabana Sarkar

To establish a successful career in Machine Learning & AI. Looking for a challenging and responsible opportunity to continually develop technical skills in software Industry. A quick learner for new initiatives, sincere and has ability to meet deadlines through effective time management, maintaining work ethics & healthy interpersonal relationships with team.Have ten years of experience in the industry, have played the role of responsible, hardworking employee.

Having good Analytical & Machine Learning skills with experience on multiple projects. Proficient in Python. Worked on Anaconda, Spyder, Jupyter. Hands on experience on NumPy, Pandas, Scipy, Scikit Learn, Github.

Hands on experience on Generalised Linear Models (Linear/Logistic/Ridge/Lasso), Decision Trees, Random forest, Grid search, ExtraTrees, kNN, SVM, Boosting Machines, XGboost, Neural Networks. Hands on experience in unsupervised learning using PCA, K-means clustering, hierarchical clustering and DBSCAN. Data visualization in Python using Seaborn. Feature engineering in Python using numpy and pandas. Missing value and outlier handling, transforming variables, creating new variables, reshaping data etc. Well versed with text mining and web scraping using Python.Have used machine learning optimization techniques like feature scaling,batch normalization,mini batch gradient descent,gradient descent with momentum,RMSProp optimization,Adam optimization,learning rate decay.

• Poject description:Pharma, A mid-sized pharma company was facing issues with counterfeits of its brands. It wanted to target and put these counterfeit operators out of business with the help of the government. In order to make maximum impact it wanted to focus on high value operators for the time being. To do that it needed to predict sales of counterfeit medicines and counterfeit operators given the data collected from various sources about their operations and sales so far so as to understand their operations size going forward.

                My Role:  Done some preliminary data analysis using pandas to find the type of data in the different columns of the given csv file.Used pipelines to process

       the data to impute missing data and convert categorical data to dummies.Created a Random Forest model to predict counterfeit medicine sales and pinpoint 

       high  value operators. 

       Link to the project code: https://github.com/shrabanasarkar/demo-func/tree/master/project3

• Manufacturing, All the money in the world can not make up for a loss of life. Better safe than sorry is the motto of all engineering and construction practices which routinely deal with dangerous situations. However many times, extent of caution is just not enough and leads to very hard learnt lessons in employee safety. Instead of relying on something going wrong and then taking preventive measure, organisations are coming up with ways of assessing the extent of danger to life for a new project before even the first hammer strikes. Given to you is a masked dataset which contains various properties of tasks taken up in heavy equipments maintenance by a manual crew. Each of these tasks have been given a hazard score which is eventually used in deciding levels of safety checks and caution while planning for the maintenance process.

      My role in the project: Input data was read from csv files. Data was preprocessed using Python pipelines. Xgboost algorithm was applied to predict the model. 

      Using    that model the mean absolute error was calculated. Using the test data, the hazards of using different tools for manufacturing were calculated.

       Link to the project code https://github.com/shrabanasarkar/demo-func/tree/master/project4

• Date of Birth: 07/18/81
• Gender: Female
• Nationality: Indian