Amogh Chaturvedi

Incoming Associate Software Engineer

2024 - Current

IT Systems and Data Analytics Intern

2024 - 2024

1.) Customer payment forecasting to reduce pending invoices: • Developed and fine-tuned a predictive ARIMA model using Python to project revenue based on the number of future customer payments based on historical payment history and current open customer invoices by performing time series analysis on data. • Created a dashboard in Power BI to show customer payment trend analysis and forecasted accuracy of model. • Integrated AWS Bedrock GenAI and RAG capability to allow data driven predictions and analysis for text-based queries sent to the model. 2.) Data analytics solution for employee survey sentiment analysis to engage and retain warehouse frontline workers: • Created employee survey data repository, cleansed the data for analysis after performing data migration. • Initiated sentiment analysis for employee onboarding data using NLTK to create vaders from chunking and Bag of words method to score how positive/negative each employee’s experience has been with the company to this point (7, 15, 30, 60, 90 days). • Performed sentiment analysis using LLMs to identify urgent worker complaints and created a dashboard for warehouse managers. 3.) Automation of customer payment exceptions: • Created a bot for cash app software automation to match customer name and payer name discrepancies and auto approve incoming payments for customer for ACH, wire, and check payments using decision tree classifiers. Efficiency increased by 80% because of bot automation. • Performed data cleaning on training data as well as supervised training for the bot in preliminary steps to improve hit rates of the model.

Undergraduate Research Intern

2023 - 2024

1.) Crack and scratch detection: • Collaborated with PhD students to conduct research in deep reinforcement learning, computer vision, and image processing to accurately detect cracks and scratches on radioactive containers, eliminating the need for physical inspections in plant factories. The image segmentation used a U-Net model to move across the image gradient and using max pooling to extract features at every given CNN layer and ultimately give an accurate prediction on whether a crack or scratch was detected. • Developed a deep reinforcement learning computer vision model in Python to differentiate between cracks and scratches on various surfaces. • Created diverse image datasets with modified parameters using Houdini image rendering software to train the model. • Implemented Python code for crack skeletonization and dilation to prevent classification based solely on image width. • Incorporated uncertainty quantification within the model to prevent over generalization of the images. 2.) Image classification: • Built a machine learning model using PyTorch to accurately identify and classify images, focusing on sceneries such as mountains, roads, bridges, valleys, and neighborhoods. • Employed random image augmentation techniques to handle variations in brightness and rotation, mitigating overfitting. • Trained the model using cross-entropy loss, utilized scikit-learn for metrics, and implemented multiclass segmentation in different convolutional layers.

Data Science Intern

2023 - 2023

1.) Supply chain optimization using algorithmic software and data: • Developed a model using Gurobi Combinatorial optimization/linear programming package to optimize supply chain planning for warehouse, product, quantity, and transit time. • Utilized CERN Root to create visualizations of supply plans for products by warehouse, providing optimal recommendations for future orders and timelines. • Integrated Docker with Ubuntu for GUI and graph display of calculations. • Trained neural networks to optimize weight passing between layers and implemented activation function adjustments during backpropagation.

Data Analytics Intern

2020 - 2021

Performed data analysis for LIVE NGO in India to analyze soil content, precipitation levels, heat, labor power, and numerous other factors to determine which crops will grow best in certain villages to provide food to those in need in malnourished and impoverished areas within Lucknow and New Delhi. This analysis helped determine which crops should be grown in which villages to maximize food stores.