I am a data scientist with a background in electrical engineering. I use my experience in electrical engineering in combination with data science to increase the efficiency of the energy systems specially renewable energies. I enjoy analyzing problems and uncovering creative solutions, whether it’s energy related or business. Through data science, I strive to develop actionable insights that can help people and companies succeed in significant ways.
• Developed a RASA chatbot using python and added new examples of intents, entities, slots and stories provided by the client.
• Tuned the hyperparameters of the NLP and Core models to optimize the performance.
• Created a dashboard in PowerBI to monitor the performance of the chatbot.
• Developed project plans, timelines and communicated with stakeholders to manage expectations and provide status updates.
• Defined metrics to track the accountability and cost of ownership for different teams within the company.
• Created data pipelines using python and SQL to generate the metrics data.
• Conducted statistical analysis and developed reports to communicate insights to stakeholders.
• Designed and developed dashboards using Unidash (internal dash tool) to visualize and present data to stakeholders.
• Thought data science to undergraduate and graduate students on topics such as python programming, SQL, data wrangling, statistics and probability, data storytelling, machine learning and big data.
• Updated existing data pipelines to collect data with more features for machine learning.
• Feature engineering and feature selection to improve the model performance.
• Developed a classification model with highly imbalanced data to detect terrorist users and improved model performance metrics.
• Created reports in Tableau to compare the performance of the old model with the updated model.
• Developed classification models using Scikit-learn and XGboost libraries
• Used Jupyter Notebook and python packages like Pandas, Numpy, Matplotlib and Seaborn for data analysis and visualization.
• Analyzed data stored in Postgres database using pgAdmin and SQL.
• Designed and developed dashboards using Tableau to present the data to stakeholders.
• Developed machine learning algorithms for the Energy Management of a U.S. house.
• Forecasted 24 hour ahead load using time series models like ARIMA, LSTM and fbprophet.
• Developed prediction models for weather situations using various machine learning algorithms.
• Provided real time estimation of the load consumption for a U.S. household.
- Explored a Kaggle dataset of home sales from 2006 to 2010 provided by the Ames’s Assessor’s office
- Used feature engineering and exploratory data analysis to combine and eliminate columns in the dataset
- Evaluated a number of models using linear regression and LASSO or Ridge regularization
- Developed a model that was able to predict housing prices with an R^2 score of %90 and RMSE of $19,000
- Scraped submissions and comments posted by Renewable and non-Renewable Energy supporters from over 40,000 subreddits using pushshift API
- Transformed data in preparation for modeling using feature engineering and natural language processing
- Identified topics that were strongly associated with Renewable and non-Renewable Energy using a logistic regression classifier
- used Twitter data, NLP, and Logistic Regression to detect the most common words people use to tweet about traffic incidents.
- Used Google Map API and Twitter API to find the route options and traffic related tweets for a specific route and a specific disaster.
- Cross reference collected Twitter posts with text features extracted from the logistic regression model to evaluate whether any of the Google map routes is likely to have hazard.
- an object detection model is developed to find solar panels from satellite images using Yolov3.
- weather API and Google API are used to obtain the weather forecast for the next five days.
- based on the object detection model and the weather forecast, solar energy production for a U.S. house is being forecast for the next five days.
- a FLASK app is created to build a web application for the project.