DEV JANI

• Scraped and preprocessed training corpora from 1,500+ websites using Python (BeautifulSoup, Scrapy, Selenium), structuring raw HTML into tokenized datasets for LLM fine-tuning • Architected a privacy-preserving federated LLM fine-tuning pipeline across decentralized data nodes using Intel OpenFL and Flower, eliminating the need for data centralization

• Designed data pipelines for preprocessing ~15,000 rows and 40+ process variables using Pandas and NumPy, ensuring data quality. • Built and deployed predictive models (Random Forest, Linear Regression) in Python and R, achieving R² = 0.94 and exposing outputs via REST-style APIs for downstream use after ensuring model reliability. • Authored structured technical documentation (~1,000 LOC in Python/script) in Confluence to ensure DevOps continuity.

• Contributed to AutoMiner, a client analytics platform enhancing data visualization pipelines and reporting systems. • Implemented advanced data visualizations (Correlation Heatmaps, Density Plots) in Python Dash, Plotly and R (ggplot2) for large-scale time-series data analysis. • Debugged and optimized reporting workflows (R Markdown with Python integration), improving performance and reliability of automated reports.