PAVANMAHAVEER SINGARA

A Passionate engineer and proactive problem solver, with 3+ years’ experience, well versed in data science concepts and machine learning models, with an objective to apply my learnings and experiential skills to contribute towards the growth of a forward-looking organization.

• Awards and Achievements: Been as a coordinator for College Event Management in Engineering Awarded first prize in Paper Presentation from (2016) State Level Technical from SIR C. R. R COLLEGE OF ENGG. On topic Motion Capture Technology. Awarded as best Class Representative and coordinator in under graduation (2017) from Department of Engineering (Sir C. R. R College of Engineering)

ML AND DL IN THE ENTERPRISE: PIPELINES AND INFRASTRUCTURE

• Open Source Contribution:
• Noticed MDTEST lacked CSV output, preventing HPC adoption. Raised issue on GitHub, discussed with maintainers and IO500 community. Opened case proposing CSV support.
• Used Determined AI platform for distributed deep learning. Faced issues setting up master services in Docker. Opened GitHub issue with configuration problems and proposed fixes.
• Reproduced and resolved software defects in GeoPandas project. Coded up fixes but did not submit pull requests.
• Pandas OpenSource Github - BUG: ExcelWriter does not retain datetime formats from set_column. FIX: Convert datetime columns to formatted strings before writing to Excel with ExcelWriter.
• BUG: pd.cut raises error with Series containing NaN when bins is an integer. FIX: Pass original list instead of Series to pd.cut to avoid error.
• Pytorch Github - BUG: Tensordot fails for int32 tensors when contracting one-element dims. FIX: Cast int32 tensors to int64/float before tensordot to avoid error.
• Capstone / Research Project- Built ARIMA model to predict house prices for Kansas government. Used time series data accounting for seasonality and trends. Generated housing market forecasts through statistical analysis. Provided data-driven insights for policymaking. Advanced capability for evidence-based decision making. Enabled data-informed approach to housing finance.
• Regression Model for Health Insurance Data( Based on USA 2021: Built multivariate regression model to predict health insurance coverage. Identified key parameters driving policy uptake. Generated data-driven insights on insurance affordability
• Image Analysis : Developed model to classify real vs fake images. Compared two approaches - custom CNN and fine-tuned VGG16. Built CNN from scratch to differentiate image classes. Fine-tuned Pretrained VGG16 via transfer learning. Evaluated and compared model performance. Applied CNN architectures for enhanced classification. Advanced image recognition capabilities. Enabled distinguishing real/fake visual data.
• Data Analysis using Python: Built GUI with Python to access university exam results. Used Pandas/NumPy to process student data. Automated lookup of scores by ID and keywords. Developed tool to retrieve and store results. Streamlined exam record management. Increased efficiency of result access/analysis. Applied Python for data processing and automation.