ASHWIN DESHPANDE
Jersey City, NJ ad2737@r.postjobfree.com 551-***-****
Linkedin GitHub
SKILLS
Programming Languages: Python, SQL
Machine learning libraries and frameworks: TensorFlow, Scikit-learn, PyTorch, Keras, Scikit-Learn Probability & Statistics: Linear Regression, Linear Classification, Logistic Regression, PCA, Support Vector Machines,Gaussian Mixture Model, Neural Networks, RNN, LSTM, CNN, GAN Tools & Development Environments: GitHub
MLOps: CI/CD, Docker, Kubernetes, Jenkins, Terraform Cloud: AWS, GCP, Azure
EDUCATION
Stevens Institute of Technology, Hoboken, NJ September 2022 - May 2024 Master of Science, Computer Science September 2022 Relevant Courses:
● Fundamentals of Computing(Python) (CS 515) : Data Structures in Python, Recursion, Object Oriented Programing, Searching and Sorting Algorithms.
● Machine Learning Fundamentals and Applications (CS 559) : Various Machine Learning Algorithms like Logistic Regression, Linear Regression, Principal Component Analysis, Support Vector Machines, Gaussian Mixture Models.
● Deep Learning (CS 583) : Neural Networks, Recurrent Neural Network(RNN), Long Short Term Memory(LSTM), Convolutional Neural Network(CNN), Generative Adversarial Network(GAN), Reinforcement Learning. University of Pune, India
Bachelor of Engineering in Electrical Engineering June 2016 - May 2020 PROJECTS
African American Bias in Hate Speech Detection Github
● Developed machine learning models using Python and scikit-learn to detect hate speech and bias against African American English in Twitter data
● Trained classifiers like SVM and Naive Bayes achieving ~80% accuracy; optimized data preprocessing and modeling to improve performance.
● Built pipeline architecture leveraging NLP techniques including tokenization, lemmatization, TF-IDF to replace biased terms and recheck classification.
● Created word embedding model with word2vec to find closest word equivalents; integrated with swear word dictionary to suggest non-offensive replacements.
● Identified over 40K mislabeled tweets out of 100K samples; able to fix classification for 26K by replacing biased terms, reducing false positives.
Real-Time Deadlift Form Tracker Github
● Developed a computer vision application to count repetitions during deadlift exercise using OpenCV and Mediapipe.
● Built a Tkinter GUI with real-time rep counting, exercise phase, and pose detection probability.
● Leveraged skills in Python, OpenCV, Mediapipe, scikit-learn, and Tkinter to deliver a polished product.
● Incorporated machine learning techniques to recognize the "deadlift" exercise, tracking the number of repetitions completed.
WORD2VEC GitHub
● Used pre-trained word2vec word embeddings to find the most similar words to the given word.
● Using the word embeddings file, which contained word2vec embeddings for 400K words and had a dimension of 50 for each vector.
● Developed a program to input a word and output the nearest words based on the word2vec embeddings.
● Made a t-sne visualization of the first 1000 words and nearest 20 words for a given word. WORK EXPERIENCE
Amazon, Bengaluru, India TRON Associate May 2021 - Aug 2021
● Worked as a TRON Associate for Amazon's Vision Operations Center, supporting machine learning and computer vision programs.
● Improved process efficiency by analyzing workflows and identifying changes to reduce handling time by 15%.
● Gained hands-on experience with full agile software development lifecycle including sprints, standups, backlogs, and kanban.
● Collaborated cross-functionally to execute projects, meeting key milestones through an agile framework. Rosenberger, Pune, India Production and Process Engineer May 2020 - Aug 2020
● As a production engineer my responsibility was to produce the input cables required for the antenna.
● In the production I was leading the team of 8 people and was given daily production targets to achieve and file reports daily for the production completed.
● I was also working with the process team for a new product which was going to go into mass production, it was a signal booster antenna used in underground metros and tunnels.