Sowmya Aitha

Senior Data Scientist with 4+ years of experience building and productionizing machine learning and Generative AI solutions for pharmaceutical, manufacturing and retail clients.

Technical

Lead for an LLM-powered customer intelligence platform processing millions of reviews, cutting LLM processing cost by ~90% and processing time by ~60%. Also delivered predictive classification (92.8% ROC-AUC HCP early-adopter model), forecasting (96% average accuracy) and recommendation systems (~$4M annual revenue impact). M.S. by Research with 3 peer-reviewed publications in deep learning and computer vision.

Senior Data Scientist at ZS Associates (2025-07 – 2026-09)

• Reduced LLM processing costs by ~90% (from $10–12K to $1–1.2K per million reviews) and improved processing speed by ~60% by migrating production NLP pipelines from text-davinci-002 to GPT-4o-mini, and by optimizing prompts and inference workflows.
• Re-architected four independent modules (Phrase Extraction, Post Processing, Parsing, Concept Collapsing) into one unified workflow, eliminating redundant processing and streamlining data exchange and execution dependencies for concept extraction, sentiment classification and evidence tagging; cut processing time and operational effort by ~70%.
• Led migration of the platform from a legacy sandbox to an AWS environment, covering architecture alignment, dependency mapping, environment planning, risk assessment and end-to-end integration testing; coordinated onboarding and knowledge transfer for ~60 team members.
• Directed technical delivery for 5–6 concurrent client pipelines processing millions of customer reviews, owning pipeline execution, debugging, root-cause analysis, output validation, production issue resolution and client-specific ad hoc requirements.

Senior Data Scientist at ZS Associates (2025-07 – 2026-09)

HCP Early Adopter Prediction Model

• Developed an HCP early-adopter classification model across multiple European markets, leveraging pre-launch engagement patterns, scientific influence and organizational characteristics to identify healthcare professionals likely to adopt newly launched products.
• Engineered 34 features capturing HCP engagement, scientific influence and organizational-level attributes; developed an XGBoost classification model with feature selection and isotonic regression for probability calibration. Implemented stratified train-test validation and 5-fold cross-validation, leveraging SHAP to identify key adoption drivers and interpret model predictions.
• Achieved 92.8% ROC-AUC and 84.8% precision in the signal-brand experiment, demonstrating strong early-adopter identification capability; leveraged SHAP to uncover key adoption drivers across brand penetration, scientific engagement and pre-launch HCP activity.

Data Scientist at ZS Associates (2022-06 – 2025-06)

Intelligent Gap Analysis

• Led the comprehensive development of a Gap Analysis module for a leading global pharmaceutical company.
• Employed a combination of NLP techniques and Generative AI to identify gaps, such as missing or extra information, between Standard Operating Procedure documents, including those in different languages.
• This module significantly enhanced efficiency of the process, saving the client considerable time, energy, and resources.

Data Scientist at ZS Associates (2022-06 – 2025-06)

Drug Price Forecasting

• Developed an eight-quarter drug price forecasting framework for a leading U.S. pharmaceutical distributor, integrating ARIMA and XGBoost to capture temporal trends and nonlinear pricing patterns.
• Achieved 96% average forecast accuracy across 118 out of 169 products, covering 92% of gross revenue.

Data Scientist at ZS Associates (2022-06 – 2025-06)

Predictive Faulty Batch Detection

• Performed comprehensive Exploratory Data Analysis (EDA) and developed a classification model to detect faulty batches during the manufacturing process before final product testing.
• Analyzed real-time sensor data from equipment and active ingredients in batch manufacturing to identify features contributing to batch failures.
• The module has been integrated into the manufacturing process, enabling early detection of faulty batches and resulting in significant time and cost savings.

Data Scientist at ZS Associates (2022-06 – 2025-06)

Personalized Customer Recommendations

• Played a crucial role in delivering personalized offers to customers of a U.S. Quick Service Restaurant (QSR) retail client.
• Leveraged QSR and customer data to generate weekly recommendations aligned with client strategies, incorporating offer recommendation, item recommendation, and customer segmentation.
• Achieved a significant improvement in conversion rates, resulting in an annual revenue increase of approximately $4 million.

M.S. by Research – International Institute of Information Technology, Hyderabad (2018-07 – 2021-12)

B.Tech in Electronics and Communication Engineering – Rajiv Gandhi University of Knowledge Technologies, Basar (2013-09 – 2017-05)