Mehul Bafna

ML/DL/AI-focused projects Automated Histological Image Segmentation • Built dual U-Net deep learning pipeline for hepatic vessel and lobule segmentation in whole-slide histological images, achieving 96.0% Dice score for lobules and 78.7% mean Dice across vascular structures through weight-boosted nnU-Net architecture • Integrated morphometric analysis (diameter, area, aspect ratio) to enable quantitative structural assessment of 1400+ liver structures Patient Readmission Risk Prediction • Developed and operationalized Random Forest model for 30-day patient readmission risk prediction (AUC-ROC: 0.88), implementing CI/CD pipeline and MLOps practices to facilitate targeted interventions and improve patient outcomes Medical Literature Summarization with Transformers • Engineered text summarization pipeline using BART Transformers for medical literature analysis • Processed 500+ research papers and reducing key findings extraction time by 25% through fine-tuned abstractive summarization Automation & workflow optimization projects • Implemented work time record management system using VBA Excel macros for date entry, time calculations, and overtime tracking, resulting in 25% improvement in data entry efficiency • Designed Python automation tool with Flask for end-to-end certificate generation, data population, PDF conversion, and email delivery, achieving 30% reduction in manual processing time

• Developed interactive Tableau dashboards to visualize cryptocurrency performance trends, leveraging Python for data cleaning, analysis, and anomaly detection • Gained experience in blockchain fundamentals (consensus mechanisms, hash functions, digital signatures) and explored the potential of Zero Knowledge Proofs for secure transactions.

- Worked on the implementation of cryptographic procedures in Python. - Hands-on experience on the computer algebra system SageMath involving combinatorics. - The usage of numpy, collections and random modules.

- Developed interactive dashboards in Tableau to visualize the performance of different encryption algorithms on financial data to compare and select the most effective algorithm. - Utilized Python to clean and analyze data related to cryptocurrencies to identify patterns and anomalies in the data, enabling researchers to detect potential threats

• Provided teaching tutorials for the 'Advanced Graph Theory and Network Algorithms' module, fostering student understanding of discrete mathematics concepts • Conducted independent research on algorithms for graph tree decompositions and domination/independence polynomials.

- Reviewed literature related to the field of research from 10 papers. - Applied potential machine learning algorithms over geospatial datasets and evaluated the results for recommending and visualizing efficient routes. - Exploratory Data Analysis and Visualization using various python modules. Programming Languages: Python, SQL, Pandas, Numpy, Matplotlib, Plotly, Seaborn

• Analyzed 10 research papers in the areas of GIS, cartography and Geoinformatics to establish a strong foundation in the research area • Applied potential machine learning algorithms over geospatial datasets and evaluated the results for recommending and visualizing efficient routes • Utilized Python modules for Exploratory Data Analysis (EDA) and visualization to facilitate data understanding.

- Presented sessions to High School and Bachelor students on how to deal with complex calculations. - Sessions were organized keeping in mind the timeframe of competitive aptitude examinations.

- Migrated data from legacy system to insurance application system using MySQL server with 98% of the data records successfully mapped - Created a database schema with 20+ tables indicating the features such as location, insurance type and liability - Addressed data modeling, structuring, and feature engineering for advanced data analysis from 15+ data sources - Created and maintained documentation, including data mappings, data dictionaries, and ETL process workflows. - Depending on business requirements, created numerous dashboards and reports using Tableau and Power BI, saving 10 hours per week of manual reporting work

- Designed and produced video lecture series for CAT preparation, including content and delivery. - Product used in 850 engineering colleges India-wide, reaching approximately 3,00,000 students. - Closely collaborated with a cross-functional team of 5 people to develop the complete CAT course. - Created power-bi dashboards for in-depth data analysis.

Completed two months of training at the IT Department, where I got exposure to the Database Management System and learned Oracle 11g architecture.