Md Shahidullah Kawsar

Experience

• 

  Sunrise Coaching Center

  Jul 2015 - Jul 2017

  Mathematics Instructor

  • Coached more than 500 high school students to their success in undergraduate enrollment tests at Bangladeshi engineering universities, especially in calculus, combinatorics, statistics, and probability• Administered student information data from 6 branches, performed data cleaning, merged different branch information, and reported to the head office every week.

• 

  Boise State University

  Aug 2017 - May 2018

  Graduate Research And Teaching Assistant

  • Award: 100% tuition fee waiver, Graduate Research, and Teaching Assistantship for 2 semesters• Mentored 100+ students for 2 semesters on homework and coding projects for Java Programming and Computer Architecture using the Open Virtual Platform (OVP)Project: Literature survey on Privacy Preserving Recommendations System using Encryption.• Identified the different types of recommendation systems on Online Social Networks.• Surveyed data privacy and data security against breaches of personal information using an efficient encryption technique while simultaneously generating the recommendation by making the original user information inaccessible to the system. Show less

• 

  University of South Florida

  Aug 2018 - May 2019

  Graduate Research And Teaching Assistant

  • Award: 100% tuition fee waiver, Graduate Research, and Teaching Assistantship for 2 semesters• Mentored 120+ students for 2 semesters on homework and coding projects for Engineering Analysis and Communication Systems using MATLABProject: Determination of Toxic Comments and Minimization of the Unintentional Model Bias• Performed Exploratory Data Analysis on the negative online behaviors, and the reason for targeting a social network user through NLP using pandas, spaCy, NLTK, Gensim, Matplotlib, Seaborn, and Plotly• Identified toxic comments from 2 million comments through Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) with Bag of Words and TF-IDF using Scikit-Learn, TensorFlow, and KerasProject: Predicting Customer Churn of a Telecom Company Using Python• Performed supervised binary classification of customer churn of a telecom company for 3500 users• Implemented Logistic Regression, K-Nearest Neighbors, Support Vector Machine, Decision Trees, Random Forest, and Xtreme Gradient Boosting to improve the model performance by 15% through cross-validation, feature engineering, grid search, and hyper-parameter tuning Show less

• 

  IDARE

  Apr 2020 - now

  Senior Data Scientist

  Project 1: AWS cloud-based IDARE Automated Machine Learning web application (IdareAI)• Leading a cross-functional team to create an AI-based solution builder tool and perform predictive analytics.• Built pipelines to Extract, Transform, and Load (ETL) big data from different sources and production-ready end-to-end ML pipelines using Pandas, Scikit-Learn, and PyTorch.• Reduced project delivery time from 12 weeks to 4 weeks through automation.• Providing active support and online training to 500 subscribers• Award: Received an innovation research grant of $200,000 from the US Department of Energy (DoE).Project 2: Visualization and Streaming Sensor Data in Dashboards• Led the development of an AWS cloud-based custom dashboard application using MS SQL Server, MongoDB, Pandas, and Plotly to visualize complex data from IoT sensors.• Implemented ETL pipelines, interactive data filtering, and drill-down features, designed the user interface, enhancing user engagement, facilitating more in-depth analysis, dynamic reporting, and generating $1M/year in revenue.Project 3: Predicting Short-term Loan Default and Customer Churn for a Bank• Handled big data from 130 branches and extracted key details about microfinance operations.• Built a pipeline to predict loan defaulters and customer churn using AutoML, and saving $500k/year.• Created financial dashboards in Power BI to get daily and monthly summaries for decision support.Project 4: Electricity Net Load Forecasting• Created a robust production-level codebase for electricity net load forecasting using Prophet and XGBoost, catering to the needs of four distinct US utility companies.• Established a comprehensive MLOps pipeline to monitor the entire ML model lifecycle, including model development, deployment, and ongoing monitoring, enabling daily scheduled predictions to support operational decisions. Show less