Shreya Tumati
Verified Expert in Engineering
Python Developer
Boston, MA, United States
Toptal member since December 5, 2025
Shreya is a generative AI engineer with experience designing and deploying production-grade AI systems across insurance, healthcare, and financial services. Focusing on end-to-end GenAI solutions, Shreya builds retrieval-augmented generation (RAG) pipelines, fine-tunes large language models, and develops autonomous agents. She also architects cloud-native AI platforms on AWS, Azure, and Google Cloud Platform (GCP) to ensure scalability, reliability, and operational efficiency.
Portfolio
Experience
- Python - 10 years
- Data Engineering - 6 years
- Artificial Intelligence (AI) - 4 years
- Google Cloud Platform (GCP) - 3 years
- Azure - 3 years
- Amazon Bedrock AgentCore - 3 years
- Amazon SageMaker - 3 years
- SQL - 3 years
Preferred Environment
Windows, Visual Studio Code (VS Code), PyCharm, JupyterLab, Slackbot, GitHub, AWS Management Console, Azure, Google Cloud Platform (GCP), Image Manipulation
The most amazing...
...solution I've built was an end-to-end, production-grade RAG and agentic AI system that transformed how a large insurance organization performed claims analysis.
Work Experience
AI Engineer
Liberty Mutual Insurance
- Delivered a large-scale RAG and agentic AI platform on AWS SageMaker that improved retrieval accuracy by 40% and reduced inference costs by 30% through optimized embeddings and dynamic orchestration.
- Built end-to-end GenAI pipelines on Vertex AI using LoRA and parameter-efficient fine-tuning (PEFT) with hybrid retrieval, significantly improving summarization precision for healthcare claims and patient correspondence.
- Deployed production-grade fraud detection and credit-risk models on Azure ML with automated CI/CD, built-in explainability using SHAP and LIME, and rigorous A/B evaluation, reducing anomaly-detection latency and strengthening regulatory transparency.
AI Engineer
Centene
- Fine-tuned models like PaLM 2 and T5 using LoRA and PEFT adapters to improve accuracy for compliance-heavy scenarios like patient summaries and medical policy interpretation, ensuring the chatbot responses met regulatory and clinical standards.
- Automated the full ML lifecycle using Vertex AI Pipelines and Cloud Dataflow, enabling continuous model updates from live claims. It supported faster deployment cycles and maintained performance KPIs while ensuring HIPAA compliance across all services.
- Designed and deployed a robust RAG system using Vertex AI and LangChain to accurately retrieve and summarize complex healthcare data, improving document search precision by 40% and enhancing decision-making speed for claims processing teams.
Data Scientist
Bank of America
- Developed and deployed real-time fraud detection models using XGBoost and BERT embeddings. These were integrated with Azure ML and SQL to flag anomalies in high-value transactions, improving fraud identification without adding latency to production.
- Migrated legacy machine learning (ML) workflows into Azure ML Pipelines and automated retraining and deployment with Azure DevOps. This ensured reproducibility, faster iteration cycles, and audit-compliant delivery across environments.
- Built transparent model-explainability tools using SHAP, LIME, and Azure InterpretML for credit risk scoring models. These outputs were used by compliance and audit teams to validate decisions, supporting Basel III and SOX regulatory reporting.
Data Analyst
Bajaj Finance
- Developed credit scoring and customer segmentation models using Amazon SageMaker, which improved targeted campaign efficiency and reduced customer churn rates by enabling real-time risk-based decisions.
- Engineered end-to-end ETL pipelines using AWS Glue and Step Functions to migrate data from on-prem SQL Server to Redshift, decreasing data processing time by over 40% and supporting scalable analytics.
- Integrated predictive model outputs into marketing and risk workflows via AWS Lambda and API Gateway, enabling near real-time upsell and retention decisions that directly improved campaign conversion rates.
Experience
Intelligent RAG-based Claims Processing System
GenAI Copilot for Medical Correspondence Automation
• Integrated LLMs (PaLM 2) with FHIR-compliant data pipelines,
• Used LangChain and Vertex AI Search for RAG-based summarization,
• Deployed APIs with Cloud Run and Vertex Endpoints for real-time service,
• Ensured HIPAA compliance using Secret Manager, Cloud KMS, and IAM.
AI Governance Dashboard for LLM Evaluation & Monitoring
• Built a custom evaluation loop using OpenAI Evals, LangSmith, and TruLens.
• Designed dashboards using BigQuery and Looker Studio to visualize drift, latency, and performance KPIs.
• Integrated guardrails and LLM-as-a-judge pipelines for safe deployment in production. It helped the business safely deploy LLMs with traceable metrics, reducing model hallucinations and building trust with compliance stakeholders.
Predictive Analytics Platform for Credit Risk and Customer Retention
Education
Bachelor's Degree in Computer Science
Malla Reddy University - Hyderabad, India
Skills
Libraries/APIs
Pandas, REST APIs, React, PyTorch, OpenAI API, Scikit-learn, Node.js, Pydantic, Cloud Key Management Service (KMS), XGBoost
Tools
Amazon SageMaker, Pytest, Microsoft Copilot, Claude, Amazon Kendra, Microsoft Power BI, PyCharm, GitHub, PaLM 2, Azure ML Studio, AWS Glue, Amazon CloudWatch
Languages
Python, SQL, TypeScript, Snowflake, BigMachines Query Language (BMQL)
Platforms
AWS Lambda, Amazon Web Services (AWS), Azure, Docker, Windows, Visual Studio Code (VS Code), Google Cloud Platform (GCP), Vertex AI, Cloud Run, LangSmith
Storage
Redis, Amazon S3 (AWS S3), Redshift
Frameworks
LangGraph, ADF, LimeJS, AWS HA
Other
Amazon Bedrock AgentCore, LangChain, Data Engineering, CI/CD Pipelines, FastAPI, Logistic Regression, Random Forests, Data Analysis, Artificial Intelligence (AI), Prompt Engineering, Retrieval-augmented Generation (RAG), Large Language Models (LLMs), OpenAI, APIs, Data Science, Machine Learning, Linear Regression, Support Vector Machines (SVM), Decision Trees, Gradient Boosting, K-means Clustering, Anthropic, Agentic AI, Vector Databases, Multimodal Models, Natural Language Processing (NLP), RESTFul APIs, Rendering, Image Manipulation, Prometheus, SAP, Deployment, Architecture, AI Assistants, FAISS, JupyterLab, Slackbot, AWS Management Console, LoRa, RAG Pipelines, Agentic RAG Systems, Pub/Sub, API Gateways, vertex endpoints, AWS Secrets Manager, Identity & Access Management (IAM), truLens, Looker Studio, Key Performance Indicators (KPIs), BERT
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring