Ketki Kinkar

Los Angeles, CA ad3bs2@r.postjobfree.com 213-***-**** LinkedIn: ketkikinkar Github: ketkikinkar EDUCATION

Master of Science in Applied Data Science Aug 2022–May 2024 University of Southern California, Los Angeles GPA: 3.92/4 Bachelor of Technology, Computer Science and Engineering Jul 2018-May 2022 MIT World Peace University, Pune, India GPA: 3.88/4 SKILLS

Programming: Python, R, C++, C Machine Learning: Scikit-Learn, Tensorflow, Pandas, Numpy, Statistical Modeling Web: HTML, CSS, JavaScript, Django, Flask Frameworks: Keras, Pytorch, Tableau, Power BI, OpenCV, PySpark, AWS, GCP Database: MySQL, MongoDB, Firebase, Hadoop Skillset: Deep Learning, Computer Vision, Data Analytics, Data Engineering WORK EXPERIENCE

Softsensor.ai, Data Science Intern, Delhi, India & US office(remote) May 2021-Nov 2021

• Built Germinal Center Cancer Detection MaskRCNN pipeline, achieving 84.8% accuracy; 0.89 specificity in lymphoma detection.

• Transformed Lymphoma detection, cutting costs from $700 to $10, reducing prediction time by 98% for clients in the US and India.

• Engineered stain normalization pipeline, ensuring 98% prediction accuracy and consistency across scanners for WSI images.

• Developed Image Quality analysis pipeline, enhancing downstream accuracy by 15%. University of Southern California, Research Assistant, Los Angeles, CA, US Jun 2023-Present

• Developing Multimodal model for Breast cancer detection, aiming to achieve model specificity of more than 0.90.

• Optimizing the harmonization pipeline to highlight tau protein on brain MRI scans, resulting in enhancement of accuracy for AD.

• Creating Alzheimer's disease detection Model to foresee disease onset 10 to 15 years in advance, with brain MRI scans, clinical data. MITWPU, Research Fellow, Pune, India Mar 2022-Jun2022

• Designed ResNet-based Transfer Learning Model for Knee Abnormality Detection with an accuracy of 88%.

• Instrumental in achieving a 45% increase in diagnostic speed and 20% enhancement in accurate injury assessment.

• Surpassed radiologist accuracy by a margin of 15% and achieved a sensitivity of 0.92. AlgoAnalytics, Data Science Intern, Pune, India Oct 2021-May2022

• Spearheaded team to create an automated Car damage Detection system with Mask R-CNN for an Insurance Company Client.

• Achieved 92% accuracy in pixel-level damage segmentation across 10,000+ images; and 82% accuracy in Classification of damage.

• Streamlined claim assessment, leading to a remarkable 65% reduction in processing time and heightened operational efficiency. PROJECTS

Financial News Research Tool Tech Stack: Python, Large Language Models, Langchain, OpenAI, Streamlit

• Developed news research tool, using OpenAI’s LLM API and Langchain framework, achieving rapid 5-second response time.

• Integrated content processing, embeddings, and FAISS library for efficient data retrieval, providing prompt-based financial insights.

• Designed an interactive user interface with Streamlit, enhancing user experience and facilitating seamless exploration of financial news data, optimizing system performance and response times contribution to informed decision-making. Loan Availability Prediction Tech Stack: Python, Scikit-Learn, Pandas, NumPy, TensorFlow, Keras

• Developed an innovative loan availability prediction model using ensemble learning techniques, achieving an accuracy of 85.8%.

• Employed AdaBoost classifier, Decision Tree, optimizing model through regularization, resulting in 7% performance improvement.

• Enhanced model's capabilities by merging ensemble learning and neural networks, contributing to 5% increase in precision. Wafer Fault Detection Tech Stack: Python, Scikit-Learn, Pandas, NumPy

• Constructed classifier automating wafer quality prediction, leading to an 80% reduction in manual testing efforts.

• Utilized K-Means for cluster prediction, dynamically employing specific models per cluster, resulting in an 86.4% accuracy.

• Enhanced sensitivity by 10% compared to standard modeling, ensuring accurate fault detection. PUBLICATIONS [20 Citations] Google Scholar

• Successfully published and presented 8 papers in distinguished platforms, including IEEE (AIST), Springer Journal, ICEARS, ICCUBEA, IJSTR, and ECS Transactions.

ACHIEVEMENTS

• Selected for publication as a Book Chapter in Springer's "Advanced Computing" volume for research, titled "The Perceived Impact of Correlative Relationship between Depression, Anxiety, and Stress among University Students".

• Received Best Research Paper Award for “Analysis of Loan Availability using Machine Learning Techniques” for IJARSCT, Sept’21

• Led team for Inter-College Smart India Hackathon (2nd Rank) to design a device for real-time monitoring of physiological and psychological parameters for military personnel.