Venkatesh Avula

At Elsevier, I lead the end-to-end evaluation of GenAI products by designing and deploying scalable, high-precision human-in-the-loop and automated assessment pipelines. This includes building rule-based engines and leveraging advanced techniques like LLM-as-judge and machine learning classifiers to measure model performance across multiple critical dimensions. I drive cross-functional collaboration with product, engineering, and leadership teams to ensure these systems deliver fast, cost-effective, and reliable evaluation at scale. My work focuses on enabling safe, high-quality AI development through robust, continuously improving feedback loops and platformized evaluation frameworks.

At Geisinger, I played a pivotal role at the intersection of healthcare and Artificial Intelligence (AI), unraveling intricate patterns within extensive datasets that included millions of electronic health records, imaging files, and textual notes. Adept at developing innovative solutions, I addressed data challenges such as missing values, skewed distributions, and imbalances while ensuring unbiased insights. My expertise lay in crafting state-of-the-art predictive models geared towards early detection of critical health conditions, significantly contributing to the evolution of precision health. Moreover, I optimized the emergency triage admissions workflow, collaborating seamlessly with interdisciplinary teams comprising physicians, research scientists, and stakeholders. Through partnerships with multiple health systems, I extended the generalizability of predictive models, operationalizing them within the healthcare system to establish a feedback loop for continuous improvement, effectively addressing model and concept drifts. Beyond model development, my commitment to advancing AI in healthcare was underscored by the publication of groundbreaking research in reputable healthcare journals, furthering the integration of AI in healthcare practices.

At Freddie Mac, I optimized the origination cycle time by implementing Optical Character Recognition (OCR) on digital documents. Additionally, I validated and improved the internal costing model for projecting loan defaults. Through segmentation analysis, I assessed the impact of changes in Loan-to-Value (LTV) buckets on various affordable lending programs. Furthermore, I investigated the influence of socioeconomic indicators on mortgage lending and refined automated valuation models for accuracy.

At GE, I led a global team in building models to predict customer responses to campaigns. Identified key customer behavior patterns using advanced clustering and segmentation algorithms. Collaborated with marketing teams to make fact-based decisions and tracked campaign performance against ROI goals.

At IBM, I conducted statistical data analysis and automated monthly performance report generation. Managed data munging tasks and developed ad-hoc reports for quantitative analysis with descriptive statistics using SAS procedures.