Chandan Chalumuri

About

Data engineering / ML practitioner with experience at MetLife building transformer-based sentiment analysis over large unstructured datasets and productionizing pipelines with Airflow/PySpark/Hadoop (reported 52% efficiency gain). Also implemented embedding-based semantic search using Pinecone/Weaviate to improve retrieval relevance and enable RAG for customer support and document matching use cases.

Experience

Education

Key Strengths

• Built and tuned transformer-based NLP models for large-scale unstructured text sentiment analysis
• Designed robust ETL pipelines (Airflow, PySpark, Hadoop) to integrate structured/unstructured data at scale
• Improved data wrangling efficiency by 52% via workflow automation and streamlined preprocessing
• Entity resolution using hybrid rule-based + ML approaches (probabilistic + fuzzy matching) with metric-driven validation
• Implemented semantic search with embeddings + vector DBs (Pinecone, Weaviate) enabling RAG in production
• Production-grade ML/data workflow practices: cloud + distributed processing + containerization + CI/CD + monitoring (MLflow/Kubeflow)
• Built and deployed production LLM assistant on Kubernetes with MLflow/Kubeflow and automated retraining
• Domain adaptation for specialized language via continued pretraining + supervised fine-tuning with negative examples
• Low-latency RAG architecture (split retrieval encoder vs smaller generation model) with caching and ANN/top-k tuning
• Production reliability practices: unit/integration testing, offline evals, canary/A-B rollouts, monitoring of latency/error and business KPIs
• Trustworthy AI delivery: explainability (SHAP/LIME) and fairness checks to drive stakeholder adoption
• Measured impact: reduced customer query response time ~43% (MetLife)
• Cross-functional delivery translating business needs into ML solutions; drove ~$2.1M incremental sales via churn/campaign optimization