Pooja Choudhary

Production ML Model Development

• Developed and deployed lead scoring ML system using Random Forest algorithm, processing 2M+ customer interactions daily with 92% accuracy, improving sales conversion by 15% and generating measurable revenue impact
• Built campaign optimization ML engine using XGBoost ensemble models with real-time inference capability (delivering 12% improvement in marketing ROI through intelligent budget allocation
• Implemented customer segmentation ML pipeline using clustering algorithms on behavioral data from 1M+ users, enabling personalized marketing campaigns with 18% increase in engagement rates

NLP & Conversational AI

• Architected NLP-based customer service bot using BERT fine-tuning and intent classification, achieving 89% accuracy on customer queries and reducing support workload by 25%
• Developed semantic search system using sentence transformers and vector databases, improving content discovery relevance by 35% over keyword-based search

MLOps & Infrastructure

• Established end-to-end MLOps pipeline with automated model training, validation, and deployment using MLflow and Databricks, reducing model deployment time from weeks to hours
• Built model monitoring framework with drift detection and performance tracking, maintaining production model reliability with 99.5% uptime SLA
• Mentored junior engineers on ML pipeline design and model evaluation best practices.

Cross-Functional Collaboration: Led technical discussions with product and business stakeholders, translating ML capabilities into actionable business strategies

Knowledge Sharing: Created internal ML engineering documentation and conducted training sessions for 15+ team members on production ML practices.

ML-Focused Data Solutions

• Engineered Customer360 ML platform using Python, Airflow, and Spark, integrating 15+ insurance data sources for ML model training, enabling 60% faster feature engineering for predictive models
• Built real-time data APIs with FastAPI and AWS Lambda serving ML applications, achieving
• Implemented data quality frameworks specifically designed for ML workflows, ensuring model training data reliability with automated validation and monitoring

Enterprise ML Infrastructure

• Designed cloud-native ML platforms on Databricks for Fortune 500 clients, building scalable infrastructure supporting ML workloads processing 100TB+ enterprise data
• Developed automated feature engineering pipelines creating ML-ready datasets from complex multi-source data, reducing model development time by 40%
• Built streaming ML infrastructure with Kafka processing 5M+ events/day, enabling real-time model inference for fraud detection and risk analytics

Healthcare Predictive Analytics

• Built predictive analytics models for healthcare claims processing using PySpark and statistical modeling, analyzing 50M+ records to identify care optimization opportunities
• Developed automated anomaly detection systems for claims data, reducing manual review workload by 35% while maintaining 94% accuracy in fraud identification
• Applied advanced statistical analysis on large healthcare datasets, providing insights that informed clinical decision-making processes

ML Data Pipeline Development

• Designed ML-ready data pipelines processing enterprise data across multiple domains, implementing feature stores and validation frameworks supporting 20+ production models
• Collaborated with data science teams on model deployment infrastructure, building robust ETL workflows that ensured model training data quality and consistency