PatentGPT: Patent Measurement Extraction System (GitHub) (GitHub - (pypi)) - (June 2023 - August 2023)

  • Developed a comprehensive patent analysis system using USPTO XML parsing and filtering, Chroma vector search, and Langchain package for patent measurement information extraction, specifically targeting patents related to Chemistry and Metallurgy.
  • Leveraged OpenAI APIs, including GPT-3.5-turbo, to perform text extraction, achieving cost-effective and efficient results. Compared various text analysis techniques such as Kor packages, Langchain text analysis, and Langchain question answering, reaching a solution that was almost 3 times faster and 23 times more cost-efficient.

AI Social Media Assistant (June 2023 - August 2023)

  • Initiated the development of a SaaS platform using Google Cloud, integrating the GPT API to function as a social media assistant. This solution is capable of generating and automatically posting content across different platforms, with current work focused on assisting Etsy stores in posting on Pinterest.

DocsGPT using Google Colaboratory (GitHub) - (April 2023 - May 2023)

  • Collaborated in the development of DocsGpt, an AI-powered app that enables anyone with a Gmail account to easily upload PDF files and ask questions using GPT-3 embedding.
  • Provided a user-friendly solution for PDF document analysis and question-answering, using Google Collaboratory, making it accessible to a wider audience without requiring installing or interacting with code

Machine Learning Course - MECE 610 (June 2021 – March 2022)

University of Alberta, Edmonton, Canada (GitHub)

  • Designed and created 50+ examples covering the fundamentals of Machine Learning, as well as Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Support Vector Machines (SVM), decision-tree, ensemble learning, and Deep Learning, with thorough theoretical explanations.
  • Developed 3 in-depth tutorials on popular Machine Learning libraries such as Scikit-learn, TensorFlow, and PyTorch, featuring practical examples in areas such as Computer Vision, Natural Language Processing (NLP), and Time Series Estimation.
  • Covering transfer learning techniques for both computer vision and NLP models to enable learners to leverage pre-trained models for their own tasks.

Deep Learning Optimal Control (January 2022 – May 2022)

Mechatronics in Mobile Propulsion (MMP), Aachen, Germany (Remote)

  • Developed a dynamics emission model using Long-Short Term Memory (LSTM) with a prediction accuracy of 96%
  • Established an imitation of the optimal controller using a deep neural network and replacing with online optimization to increase optimization speed by 50x faster

Real-time Vehicle Emission Model (January 2021 – December 2021)

IAV GmbH, Gifhorn, Germany (Remote)

  • Automated a systematic method using K-means clustering algorithm to select the most accurate ML model based on application of model to decrease development time up to 8 times

Build Deep Learning Models with TensorFlow Projects (November 2021)

  • Built deep learning classifier using TensorFlow with Keras to predict forest cover type based only on cartographic variables (GitHub)
  • CNN-based classification of Covid-19 and Pneumonia based on X-ray lung scans using TensorFlow with Keras (GitHub)

IBM Applied Data Science Capstone project: Edmonton’s Best neighborhood (March 2021) (GitHub)

  • Deployed web scraping using beautiful soup package of Python to collect Neighborhood name, postal code, and locations
  • Employed Foursquare API to mine features of Edmonton’s neighborhood and deployer K-means clustering algorithm

Technical report: AI and MPC Applications for Engine Control and Calibration (November 2020 – February 2021)

Cummins, Columbus, United States (Remote)

  • Analyzed state-of-the-art literature in the field of AI and MPC for automotive applications in collaboration with Cummins
  • Critical reviews of existing methods for implementing real-time adaptive learning
  • Identified promising ML methods to address automotive industry challenges