NIKITA GOSWAMI

+1-609-***-**** adid4b@r.postjobfree.com linkedin.com/in/nikita-goswami github.com/nikita018 EDUCATION

University at Buffalo, State University of New York GPA: 3.85/4.00

Master of Science in Data Science (Graduating in Dec 2020)

Machine Learning, Computer Vision, Databases, Algorithms, Numerical Mathematics, Probability and Statistics, Data Mining Banasthali University, India

Bachelor of Engineering (Algorithms, Data Structures and Object Oriented Programming) (Jun 2011 - May 2015)

TECHNICAL SKILLS

Languages: Python (NumPy, Pandas, Matplotlib, Tensor Flow, Keras, Flask), R, MATLAB, JavaScript, Java, XML, Hadoop Analytical Skills: SQL, Hypothesis Testing, A/B testing, EDA, Tableau, Regression, Excel, MongoDB, Time series analysis, ARIMA, 2SLS Machine Learning: KNN, Logistic Regression, SVM, Decision Trees, Dimensionality Reduction, Neural Networks (CNN, LSTM, RNN), Unsupervised Learning, Feature Engineering, Hyper parameter optimization, Market basket Analysis, Clustering, Anomaly Detection Online Courses: Coursera - Machine Learning, Deep Learning specializations – Andrew NG, Algorithms (Stanford University) PROFFESIONAL EXPERIENCE

Artificial Intelligence Intern, MTX Inc (June 2020 - Present)

Built a customizable Recommendation Engine product using collaborative and content-based filtering algorithms

Identified tampered regions in a document by defining it as a semantic segmentation problem and using U-Net architecture with 88% accuracy. Applied OCR techniques to identify text and experimented with texture descriptors and color histograms

Developed a decision board for MLB team managers and players to recommend the action that returns maximum situational advantage. Clustered and profiled similar players based on their diverse action to different situations Data Scientist, [24]7.ai India (Apr 2017 - Jun 2019)

Implemented Support Vector Machine based NLP intent prediction model to enable automated responses in an IVR and chatbot system for an insurance client. The bot helped reduce the human efforts by 15%, saving $150 k annually

Deployed customer targeting model based on incremental propensity to convert given a proactive chat option on the web, drove $ 3 million incremental revenue annually. Forecasted number of chat agents catering to the demand

Identified the root causes for call failure through data analysis, performed threshold optimization, synthetic data introduction, improved model generalization resulting in a growth of 5% call containment in the automated IVR system Data Analyst, Tata Consultancy Services India (Jun 2015 - Apr 2017)

Identified potential consumer base for multiple marketing campaigns, engineered leading indicators and improved customer targeting with personalization, drove $ 1.5 Million incremental revenue annually

Built Natural Language based customer interaction capability in IVR systems on top of the conventional choice- based interaction. Performed hyper parameter tuning on the Neural Network classifier to improve accuracy by 2%

Designed and analyzed A/B test experiments on greeting messages design for an IVR system to measure effectiveness. The results were analyzed in R using Hypothesis testing. Mentored a team of 3 to enable them to work independently on analysis

ACADEMIC PROJECTS AND RESEARCH

Web Scraping and Topic Modelling

Key Tech Components: NLTK, sklearn, Lemmatization, Latent Dirichlet Allocation.TF-IDF feature matrix, Ngrams, SVD Scraped web data through API to study the trending topics on a day. Cleaned the data, performed feature engineering using TF-IDF feature matrix including ngrams and used SVD to cluster the topics

Medical Diagnosis – Chest X-Ray Diagnosis with Deep Learning Key Tech Components: TensorFlow, keras, Image Augmentation, Transfer Learning, Hyperparameter Tuning Prepared the data, handled class imbalance using a weighted loss function, used transfer learning to retrain DenseNet model, measured diagnostic performance using AUC-ROC curve and visualized model activity using GradCAM

Portrait Mode – Depth effect in pictures using Semantic Segmentation Key Tech Components: TensorFlow, Keras, cv2, matplotlib, Mobilenet weights, U-Net Architecture, Convolution Filters Used computer vision, Deep learning along with transfer learning to find out foreground and background information in an image. Sharpened the main object and applied gaussian filter to blur the background to create depth effect in pictures

Business Buddy – Actionable Insights from Yelp reviews Key Tech Components: Numpy, Pandas, matplotlib, plotly, seaborn, SQLite, Logistic Regression Extracted features from user reviews, fitted a logistic regression model for sentiment analysis with 90% accuracy

Path Navigator – Decision making for self-driving cars Worked on a module to predict presence of a road (area fit for driving a car) in front of a vehicle

Contact this candidate