Ruchitha Paccha

AI/ML Engineer with 5+ years of end-to-end experience designing, building, and deploying production-grade Generative AI solutions including LLM-powered features, retrieval-augmented generation (RAG), and agentic workflows to support business and product goals. Built LLM-powered summarization, extraction, and classification features using GPT-4, BERT, and PaLM in production enterprise environments; designed RAG pipelines covering document ingestion, chunking, embeddings, vector search, and prompt engineering using LangChain, LlamaIndex, FAISS, ChromaDB, Pinecone, and Azure AI Search. Developed agentic workflows with guardrails grounding, citations, and refusal behavior for safe, reliable AI in regulated healthcare and enterprise settings; created offline and online GenAI evaluation frameworks using test sets, automated metrics, and human review processes tracked in MLflow.

Expert

Python programmer with clean, maintainable, well-documented code; experienced in cross-functional collaboration with data scientists, software engineers, security, and product teams across AWS, Azure, and GCP cloud environments.

AI Engineer at WellCare Medicare (2025-06 – Present)

• Built and iterated on LLM-powered features summarization, extraction, and classification using GPT-4 and PaLM on Azure ML, deploying production-grade Generative AI solutions that automated Medicare document workflows, increased throughput by 35%, and served 500K+ monthly users at 99.9% uptime.
• Designed end-to-end RAG pipelines (document ingestion, chunking, embeddings, vector search, prompt templates) using LangChain, FAISS, ChromaDB, and SentenceTransformers, building real-time claims Q&A systems for clinical staff that achieved 91% query accuracy grounded in clinical document context.
• Developed agentic workflows with guardrails grounding, citations, and refusal behavior using LangChain and LangGraph, implementing safety controls for regulated Medicare AI applications compliant with HIPAA and CMS requirements.
• Created offline and online GenAI evaluation frameworks using test sets, automated quality metrics, latency monitoring, and human review processes tracked in MLflow, enabling continuous quality improvement across production LLM deployments.
• Built reproducible data pipelines for model inputs and evaluation using Azure Data Factory and Python; deployed and optimized GenAI services for latency, token usage, caching, drift detection, and incident response on Azure ML.
• Collaborated cross-functionally with data scientists, software engineers, security, and product teams to translate Medicare business requirements into safe, scalable GenAI solutions; contributed to code reviews, runbooks, and user documentation

Research Assistant at University of South Florida (2023-08 – 2025-05)

• Designed RAG solutions across 5 production research AI systems using LangChain, LlamaIndex, FAISS, ChromaDB, and PGVector, enabling 30+ researchers to query 50M+ records via natural language with 30% improvement in retrieval accuracy over keyword baselines.
• Built LLM-powered summarization, extraction, and classification features using domain fine-tuned transformer models; developed agentic workflows with guardrails for grounding and citation enforcement; contributed findings to 2 IEEE peer-reviewed publications on safe GenAI in real-world settings.
• Created offline and online GenAI evaluation frameworks using golden test sets, automated metrics, qualitative rubrics, and human review processes tracked in MLflow and Weights and Biases, ensuring reproducible, measurable quality improvements across all research models.
• Built reproducible data pipelines using Azure Data Factory and Python; assisted with training and evaluating traditional ML models for classification, regression, anomaly detection, and recommendation across healthcare research datasets with full HIPAA, IRB, and NIST 800-53 compliance.

AI/ML Engineer at Wipro – Estée Lauder Companies (2021-08 – 2023-06)

• Built LLM-powered features using BERT, DistilBERT, and GPT-equivalent models via Hugging Face, deploying production-grade GenAI and ML solutions across 35 global markets for 20M+ users improving product review classification accuracy by 22% and recommendation click-through rates by 18%.
• Designed RAG pipelines using SentenceTransformers, FAISS, and LangChain for context-aware product search and catalog retrieval; deployed and optimized GenAI services for latency, token usage, caching, and drift detection via SageMaker Pipelines and CloudWatch.
• Implemented offline and online GenAI evaluation frameworks using MLflow and SageMaker Experiments, reducing model failure rates by 60%+ through systematic quality, safety, latency, and cost evaluation across production deployments.
• Built reproducible ETL/ELT data pipelines using Kafka, AWS Glue, and Airflow; developed traditional ML models for classification, regression, anomaly detection, and recommendation; maintained full CI/CD with Terraform and GitHub Actions.

Data Scientist Intern at Elevance Health (2020-06 – 2021-07)

• Built LLM-powered extraction and classification features using AWS Comprehend Medical and ClinicalBERT for ICD code extraction, increasing accuracy by 25%; designed RAG-style retrieval pipelines for clinical document search.
• Implemented offline evaluation frameworks using golden test sets, automated metrics, and SHAP-based human review processes, increasing predictive model adoption by 45%; developed traditional ML models on 2M+ monthly records with full HIPAA/GDPR compliance.
• Built reproducible data pipelines using AWS Glue, Lambda, and Step Functions; wrote clean, well-documented Python across all GenAI and ML deployments.

M.S. in Business Analytics and Information Systems – University of South Florida (2023-08 – 2025-05)

B.Tech. in Electronics and Communication Engineering – Kalasalingam University (2017-07 – 2021-07)