Yashodhar Pansuriya

Master Thesis at Boehringer Ingelheim

Built an end-to-end ML pipeline to predict IPF disease progression using clinical data and HRCT radiomics. Performed EDA, preprocessed HRCT scans with automated lung segmentation, applied feature selection to reduce high-dimensional radiomics, and trained and compared Linear, Lasso, Elastic Net, PLS, and Random Forest models; interpreted results with SHAP and PDP/ICE.

Intern at Boehringer Ingelheim

Developed Carlos, an AI tool to extract data from lengthy clinical trial reports; designed methodology to extract and standardize tables into clean JSON, and convert into XML for integration; added real-time review/editing features to ensure data integrity in clinical trial reporting.

Working Student - Software Engineer at Fraunhofer Institute for Silicon Technology

Developed and enhanced MES software for silicon chip production; contributed to core business logic and feature improvements; wrote unit and integration tests; analyzed machine data to extract production metrics; improved UI and REST APIs for data exchange with production machines.

Project Intern: AI Developer at Kubermatic

Contributed to CNCF LLM project to simplify the Cloud Native ecosystem with context-aware answers; built data pipeline to collect and standardize GitHub and Stack Overflow data into Q&A pairs; trained transformer-based models on HPC clusters and Hetzner servers for scalability; integrated a Local AI Chat UI, deployed it on Hetzner, and automated deployment with GitHub Actions.

Software Engineer at Honest Software Solution

Built data mining capabilities to fetch data from GitHub and Jira; designed schemas for multi-source data and stored data in PostgreSQL; created KPI dashboards for monitoring across sources.

Master Thesis (Radiomics Analysis and Machine Learning Modeling for Prognostic Biomarkers in Idiopathic Pulmonary Fibrosis) at Boehringer Ingelheim

Built an end-to-end ML pipeline to predict IPF disease progression using clinical data and HRCT radiomics; conducted exploratory data analysis; preprocessed HRCT scans and performed automated lung segmentation for consistent radiomics extraction; applied feature selection to reduce high-dimensional radiomics and retain informative predictors; trained and compared Linear, Lasso, Elastic Net, PLS, and Random Forest models and interpreted results with SHAP and PDP/ICE.

Clinical Data Science Internship (Bayesian Statistics) at Boehringer Ingelheim

Built 'Carlos', an AI tool to extract data from lengthy clinical trial reports; developed methodology to extract tables and convert formats into clean JSON; standardized extracted data in chunks and converted to XML for seamless integration; implemented real-time review and editing features to ensure high data integrity in clinical trial reporting.