Hire Singamsetty S. - Generative AI Architect

* Designed and implemented end-to-end architectures for Generative AI applications, ensuring scalability, security, and performance optimization. * Integrated large language models into existing enterprise platforms with seamless interoperability. * Fine-tuned pre-trained foundation models for domain-specific use cases using structured data collection, sanitization, and preparation strategies. * Evaluated and deployed GenAI frameworks such as PyTorch, TensorFlow, LangGraph, CrewAI, AutoGen, Agentic workflows, and AgentFlow. * Defined and executed AI transformation roadmaps by identifying high-impact GenAI use cases. * Led and mentored cross-functional teams including data scientists, GenAI engineers, DevOps, and software developers. * Monitored and optimized model performance, infrastructure utilization, and computational costs for large-scale deployments. * Ensured responsible AI practices by implementing governance, data privacy, bias mitigation, and ethical AI safeguards. * Collaborated closely with business stakeholders to translate complex requirements into effective AI solutions.

Project 1: Generative AI-Based Automation for ServiceNow Incident Resolution Designed and deployed an end-to-end Generative AI solution to automate access-related incident resolution in ServiceNow. Converted historical data from Snowflake into embeddings using Azure Models, stored in Milvus (IVF-Flat), and applied Cosine Similarity for contextual incident matching. Built a GPT-based classification system to separate access vs. non-access tickets and sub-categorize IAM cases. Developed a RAG pipeline using LangChain + OpenAI GPT to generate automated resolution steps. Integrated with Ignio bot to auto-update tickets and trigger resolution emails. Applied fine-tuning with real-time feedback and RLHF to improve model accuracy and reduce drift. Containerized with Docker and implemented comprehensive testing to ensure scalability and stability. Project 2: Automated Insurance Claim Analysis using LLMs Automated claim analysis and decision-making using LLMs and similarity search. Trained models on insurance handbooks and policy docs for contextual claim understanding. Ingested and validated datasets (customer, policy, medical records) via LangChain pipelines and applied automated EDA for trend detection. Generated embeddings with Word2Vec, stored in a vector DB, and used Cosine Similarity for claim-policy matching. Leveraged OpenAI GPT to assess claim status, detect fraud, and generate structured settlement reports. Integrated Hugging Face Transformers to benchmark model performance on medical text. Built a RAG pipeline combining vector DB retrieval with LLM reasoning. Parsed LLM outputs to extract insights and deliver concise claim summaries.

Project: Automated Data Observability System Responsibilities: Developed a real-time data observability system integrating Azure Databricks, Azure Data Factory, and Apache Kafka for pipeline monitoring. Implemented API-based parsing to extract log metadata and stored insights in Azure SQL DB. Automated alerts for failures and SLA breaches via Microsoft Graph API, MS Teams, and ServiceNow. Streamed pipeline metadata into Prometheus and visualized metrics on Grafana dashboards. Used Docker and GitHub CI/CD pipelines for scalable deployment. Conducted unit testing to ensure reliability of ETL scripts and API services. Applied Scikit-learn for anomaly detection in pipelines. Built custom operational visualizations using Matplotlib and Seaborn. Followed SDLC practices to manage ML models and data workflows in production.