kasula sandeep

At Deloitte, I architect and develop enterprise document ingestion and knowledge management platforms for Philip Morris International. I build Python microservices on AWS ECS Fargate, orchestrating parsing, PII redaction, semantic chunking, and vector embedding workflows.

I've built RAG-based conversational search with OpenSearch and PostgreSQL pgvector, supporting hybrid retrieval, real-time chat streaming, multi-tenant APIs, and Cognito-based authentication. I also developed DemandForest, a cold-start forecasting service that predicts 12-month market-share trajectories for product launches.

Previously at Zebra Technologies, I integrated Gemini and transformer-based LLMs for semantic search, text generation, and summarization, increasing automation capabilities across enterprise tools by 35%. I also deployed XGBoost models through Vertex AI and Kubeflow pipelines and built REST APIs for real-time inference.

I work across Python, AWS, GCP, machine learning, NLP, and MLOps to turn raw enterprise data into production AI systems. My experience spans document intelligence, predictive modeling, cloud-native architecture, and generative AI applications.