Kush Suryavanshi

Boston,MA 917-***-**** ad3oqf@r.postjobfree.com LinkdeIn Availablity: May 2024 onwards EDUCATION

Northeastern University, Boston, MA September 2022 - May 2024 Masters of Science in Artificial Intelligence GPA - 3.6 TECHNICAL SKILLS

Languages: C++, Python, MATLAB, SQL

Frameworks & Libraries: Tensorflow, Pytorch, scikit - Learn, OpenCV, Pandas, NumPy, Matplotlib, Seaborn Skills: Statistics, Machine learning, Deep Learning, Data Analytics, Natural Language Processing, Computer Vision WORK EXPERIENCE

BARC India, Mumbai, India July 2023- August 2023

Data Science Intern

● Conducted in-depth analysis of demographic survey data, applying advanced statistical techniques and machine learning models to identify key audience behaviors and preferences, resulting in a 20% improvement in targeting accuracy.

● Devised and implemented innovative algorithms to process and analyze data, reducing processing time by 30% and enhancing the accuracy of results by 15%.

● Collaborated with a dynamic team at Barc India to transform raw data into meaningful insights, contributing to the development of actionable recommendations for strategic planning initiatives. RadicalX, New York, New York May 2023- June 2023

Artificial Intelligence Intern

● Collaborated with the development team to design and implement AI-powered features and functionalities of the platform.

● Assisted in improving and refining the AI Dev Manager powered by GPT-3, utilizing natural language processing, machine learning, and deep learning technique that drove personalized recommendations and tailored learning paths for users. Indian Institute of Technology, Mumbai, India April 2021- November 2021 Machine learning Intern

● Developed a speech-to-text system utilizing Google Cloud, Pytorch, and Wavenet to convert voice recordings into text with 95% accuracy.

● Successfully implemented a neural voice cloning system that learned to synthesize audio, showcasing a 50% improvement in recognition rate.

PROJECT EXPERIENCE

Anomaly Detection in Chest X-ray Using Autoencoder Network

● Developed and implemented a novel anomaly detection method using dual-distribution discrepancy, resulting in a 14.6% improvement over the AE baseline and 10.8% over the MemAE baseline on RSNA Pneumonia Detection Dataset.

● Utilized inter- and intra-discrepancy scores to identify anomalies in CXR datasets, achieving a 4.3% improvement over AE-U baseline on the VinBigData Chest X-ray Abnormalities Detection dataset.

● Conducted experiments on RSNA Pneumonia Detection Challenge and VinBigData Chest X-ray Abnormalities Detection, which resulted in the successful implementation of our anomaly detection method to improve performance by up to 14.6%. Hand Gesture Recognition using 3D CNN and Recurrent Neural Network

● Developed a 3D CNN architecture to extract spatiotemporal features from video sequences and used a combination of 2D CNN and LSTM to extract spatial features and learn the temporal dependencies between frames and improve the accuracy of the model. Regularized using GridSearchCV and achieved a further improvement in accuracy to 94%.

● Preprocessed the dataset by resizing images to a consistent size of 100x100 and normalizing pixel values to have zero mean and unit variance. Trained the model on a dataset of 700 videos, achieving an accuracy of 92% on the test set. Multimodal Emotion Recognition using Recurrent Neural Networks

● Developed hybrid models of BiLSTM and Dialogue RNN architectures on MELD Dataset for emotion classification achieving 67% maximum accuracy with a team of three.

● Utilized various techniques like Principal Component Analysis, signal processing, and Deep Learning to analyse multimodal data from various platforms.

● Tested 6 different modality combinations on a 5-fold cross-validation sample set resulting in an average precision of 62%.