Dvija Shah

About Dvija Shah

I’m a Data Scientist focused on transforming complex data into intelligent, scalable solutions that drive measurable business outcomes. With over 2 years of experience applying Machine Learning, NLP, and Generative AI in production, I specialize in experimentation, causal analysis, and model-driven decision-making that accelerate product and user growth.At WebMD (Medscape), I’ve led AI and data science projects from experimentation frameworks to Generative AI pipelines, impacting millions of healthcare professionals globally.Key Highlights:• Conducted A/B tests and causal inference analyses across 5M+ users, increasing targeting efficiency and UCR by 3%.• Built LLM-powered content generation workflows (GPT-4, Claude, Gemini), reducing editorial workload by 50%.• Developed claims-based audience scoring models, boosting campaign reach by 40%.• Fine-tuned NLP models and engineered a multi-agent pipeline in Dify, reducing manual tagging by 65%.• Automated ETL pipelines in Databricks and SQL for 100M+ records, enabling faster insights and dashboards.I thrive at the intersection of data science and AI innovation. My focus is on designing data-driven systems that combine statistical rigor, experimentation, and automation to deliver insights and products that create measurable business value.

Experience

1. Data Scientist

   WebMD

   Sep 2023 — Present

   Data Scientist at WebMD (Internet Brands)Enhanced Audience Targeting: Developed a dynamic claims scoring algorithm, increasing email targeting reach by 40% and boosting user engagement with personalized content.Improved Predictive Models: Enhanced scoring accuracy for a nationwide SSDI firm, increasing prediction accuracy by 20% through advanced feature engineering and weighted classification using Patient Data and Rx Data.Designed and Deployed AI Solutions: Built a generative AI tool leveraging GPT-4, Claude-3.5, and Gemini, streamlining content creation workflows and reducing editorial workload by over 50%.NLP-Focused Innovation: Fine-tuned Bio-BERT models for clinical article classification, achieving 82.65% accuracy to optimize content for expert-level physicians.Optimized Omni-Channel Engagement: Designed and executed an omni-channel engagement strategy for healthcare professionals, improving cross-channel engagement by 30%.Collaborated Cross-Functionally: Partnered with engineers, data scientists, and product managers to create scalable solutions aligned with strategic business goals.

Education

• L J Institute of Engineering and Technology (LJIET)

  Computer Engineering, Computer Science

  2015 — 2019

• Northeastern University

  MS in Data Science

  2021 — 2023