Open to opportunities

Rushabh Shah

@rushabhshah3

Data Scientist at North Carolina State University who deployed a RAG system that cut information retrieval time by 50%.

At North Carolina State University, I deployed a production-grade RAG system for searching codebases, PDFs, and technical documentation, reducing information retrieval time by 50%. I also engineered a data pipeline with more than 850,000 records and developed an environmental risk classification model.

At Schneider Electric, I architected an LLM chatbot that saved 3 to 4 hours daily. Earlier, as an AI Engineer at DXC Technology, I developed NLP document processing and predictive analytics solutions. My projects include ClinicFlow, an AI clinical documentation platform, and RestaurantVoiceHub, an AI voice ordering assistant.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist

Current

North Carolina State University

Feb 2024 - Present (2 years 8 months)

Deployed a production-grade RAG system using FAISS, Gemini API, Google Cloud Vertex AI, and LangChain for intelligent semantic search, reducing information retrieval time by 50%. Engineered an 850K+ record data pipeline using Apache Airflow and Spark, and generated an ensemble ML solution achieving AUC-ROC of 0.93.

Data Science Intern

Schneider Electric

May 2024 - Aug 2024 (3 months)

Architected and migrated a custom LLM chatbot using AWS Bedrock (Claude3.5 Sonnet), AWS SageMaker, and prompt engineering, saving 3 to 4 hours daily and achieving 90% response accuracy. Optimized field service training schedules by developing a DBSCAN-based clustering algorithm in NumPy.

AWS Claude Sonnet 3.5 AWS Sagemaker Prompt Engineering DBSCAN NumPy S3 Amazon Redshift

AI Engineer

DXC Technology

Jan 2022 - Jun 2023 (1 year 5 months)

Prepared and analyzed structured datasets using MySQL, Python (SciPy), and TensorFlow, applying regression modeling, time series forecasting, and causal inference to improve prediction accuracy by 25%. Designed NLP-based document classification and information extraction solutions using Hugging Face Transformers and spaCy.