• Phone: +1 980-899-3900
• Email: kunaljshah03@.gmail.com
• LinkedIn: linkedin.com/in/kunal-shah-86a23016b
• GitHub Profile: github.com/kunalshah03
• Google Scholar Profile

Education

• Master of Science: Computer Science (GPA: 3.75/4.0)

  • North Carolina State University (08/2022 - 05/2024)
  • Relevant Coursework: Neural Networks and Deep Learning, Database Management Systems, Design & Analysis of Algorithms.

• Bachelor of Engineering: Computer Engineering with Honors: Data Science (GPA: 9.66/10)

  • Pune Institute of Computer Technology (08/2018 - 06/2022)

Skills

• Machine Learning Deep Learning, Computer Vision, Natural Language Processing, Generative AI, Prompt Engineering.
• Languages Python, C++, SQL, Javascript, Java.
• Frameworks PyTorch, TensorFlow, Keras, Flask, Django, NodeJs, ReactJs, Spring Boot.
• Cloud, Databases & Tools AWS, Google Cloud Platform(GCP), MySQL, MongoDB, Git, Tableau, Docker, Microsoft Office.

Experience

Research Assistant @ North Carolina State University - Raleigh, NC | 01/2023 - Present

• Developed a transferable prediction model, leveraging hierarchical clustering with CARTs and transfer learning, to assess open-source(OS) project health (95% accuracy), analyzing 150+ GitHub projects with interpretable insights from LIME.
• Optimized regression models for OS health metrics using HyperOpt, reducing mean squared error by 40%.
• Contributed to open-source tools for automated data mining in OS projects and participated in NSF-funded research on adapting software engineering practices to computational science.
• NSF Award

Data Science Intern @ Fidelity Investments - Durham, NC | 06/2023 - 08/2023

• Built and deployed a cutting-edge Generative AI text-to-image model using AWS SageMaker for the Fidelity platform.
• Evaluated potential Fidelity software for automated data analysis on a 2.1M-participant dataset using EDA, feature engineering, and advanced data visualization.
• Presented evaluations and recommendations to senior management, impacting decisions on software adoption.

Machine Learning Engineer Intern @ Omdena - Pune, India | 05/2021 - 07/2021

• Implemented user-driven NLP solutions reducing information-gathering time by 70%.
• Employed LSTM/BiLSTM/Transformers for 96% accurate data extraction from diverse sources with sentiment analysis and summarization.
• Deployed the web-scraping tool on the Google Cloud Platform with a Streamlit front-end.

Software Developer Intern @ Carikture India - Pune, India | 07/2020 - 09/2020

• Launched a data-driven educational platform (Node.js, Electron.js, React.js) for 5000+ users, integrating predictive models and interactive data dashboards (Tableau) to personalize learning paths, increasing user engagement by 40%.
• Incorporated machine learning models for personalized learning paths and A/B tested UI elements to optimize the user experience.

Publications

• EffResUNet: Encoder-decoder architecture for cloud-type segmentation

  • Researched a cloud-type segmentation encoder-decoder UNet-architecture leveraging EfficientNet, residual blocks, and an attention mechanism with an F1 score of 0.5735, surpassing SOTA methods.
  • Published in Big Data and Cognitive Computing 2022.

• Image-Based Plant Seedling Classification Using Ensemble Learning

  • Devised an ensemble-based model for plant seedling classification using a shallow CNN for feature extraction and K-Nearest Neighbors for classification refined through 5-fold cross-validation. Achieved a recall of 95.32%, outmatching existing techniques.
  • Published in ICACIE 2021 (LNNS volume 428, Springer).

Projects

• Pose Detection for Sign Language Videos

  • Programmed a deep learning pipeline for sign language-to-Hamnosys translation, achieving 47% accuracy (9% above SOTA) using 3D pose landmarks from Mediapipe and transfer learning on TensorFlow CNNs, enabling multilingual communication.
  • Technologies: Python, Tensorflow, Mediapipe, OpenPose, Deep Neural Networks

• Real-time Video Anomaly Detection

  • Created a multi-faceted deployable tool integrating real-time traffic visualization, pedestrian anomaly detection system (ROC AUC 0.932, 20% false positive reduction), and a Streamlit-powered alert system for proactive notifications to enhance traffic safety and security.
  • Technologies: PyTorch, Python(Pandas, NumPy, Plotly ), OpenCV, YOLO, CNN, Twilio, SendGrid

• WolfMedia

  • Designed and built a scalable media streaming platform (WolfMedia) using Spring Boot and Docker. Managed artists, content, payments, and subscriptions with secure APIs and comprehensive documentation.
  • Technologies: Java, Spring Boot, SQL, Docker, Swagger API

• PackTravel

  • Developed a ride-sharing application (PackTravel) using Django and MongoDB for university students. Enabled features like multi-modal route planning, ride search/joining, and user authentication. Planned future ML-powered route optimization and cab service integration.
  • Technologies: Python, JavaScript, Django, MongoDB, Pylint

• PacReviewVersion2

  • Enhanced an existing Jobs Review System application, enabling users to share anonymous job reviews, featuring comprehensive filtering, pagination, and MongoDB migration for scalability and anonymity, leveraging Python, Flask, JavaScript, SQL, MongoDB, Git, HTML5, CSS, and Bootstrap for implementation.
  • Technologies: Python(Flask), JavaScript, SQL, MongoDB, Git, HTML5, CSS, Bootstrap

Leadership & Initiatives

• Engaged as a research member at the NCSU RAISE Lab, focusing on the convergence of Artificial Intelligence (AI) and Software Engineering (SE).
• Contributed as a technical consultant for 3 startups incubated at the PICT Startup and Innovation Center as the technical head of the PICT Entrepreneurship Development Cell.
• Led seminars on Machine Learning for 500 students as a recognized PICT ACM Student Chapter (PASC) member and contributed to 3 Best ACM Student Chapter awards.