Laya Avula

About Laya Avula

AI/ML Engineer with 3+ years of experience building scalable machine learning and Generative AI systems across healthcare and financial services. I specialize in developing production-grade AI solutions that combine machine learning, large language models (LLMs), and modern data engineering.My work focuses on designing end-to-end AI systems—from data pipelines and model training to API deployment and monitoring on cloud platforms such as AWS, Azure, and GCP. I have built real-time ML services, RAG-based chatbots, document intelligence pipelines, and multimodal AI systems that improve automation, decision-making, and operational efficiency.Currently at New York Life, I develop LLM-powered applications, scalable APIs, and cloud-based ML systems that support real-time applications and intelligent automation. My work includes optimizing model performance, reducing LLM costs, implementing responsible AI practices, and deploying production-grade pipelines using tools like FastAPI, Docker, SageMaker, and vector databases.I enjoy solving complex data problems and turning advanced AI capabilities into practical, reliable products that deliver measurable impact.Key Interests: Generative AI, LLM Systems, Retrieval-Augmented Generation (RAG), AI Agents, MLOps, and Scalable Machine Learning Infrastructure.

Experience

1. Ai Ml Engineer

   New York Life

   Feb 2025 — Present · US

   Built and deployed REST APIs using FastAPI, Docker, and AWS Elastic Beanstalk to expose machine learning services for client-facing real-time applications, achieving sub-100ms latency.• Deployed scalable machine learning models on AWS SageMaker and EC2 instances for production inference, setting up endpoint monitoring and performance alerts using Amazon CloudWatch.• Developed and deployed AI models using Llama 8B and Mistral 8B, optimizing real-time text summarization and automation.• Built Summarizer V1 & V2 with a chat-based interface, improving text summarization accuracy and enhancing data visualization.• Built a production-grade RAG chatbot using Python, deployed on AWS, integrating SQL, vector databases (Pinecone, FAISS), and knowledge graphs for real-time, context-aware query handling.• Increased document processing efficiency by 75% by building OCR pipelines using Tesseract, AWS Textract for accurate text extraction from documents.• Reduced LLM response time by 55% via prompt engineering (Chain-of-Thought, Few-Shot) and memory management, increasing user engagement by 30%.• Implemented security guardrails and input validation to defend against prompt injection attacks (security attacks); added logging and unit tests for LLM decision flow.• Reduced GPT-4o API operational costs by 20% through token optimization, batching, and request deduplication, ensuring scalable performance under high user load.• Developed a Large Language Model (LLM) system EndoAssistant, for image-based Q&A system in medical surgery, achieving 93% contextual accuracy by fine-tuning LLaMA and CLIP.• Built a scalable RAG pipeline using Autogen and vector databases, integrating fine-tuned LLMs to deliver grounded and contextaware answers.• Automated model retraining workflows by integrating AWS Glue for data refresh, SageMaker Pipelines for model retraining, and AWS Lambda triggers.• Designed a speech-to-text pipeline combining Whisper with GenAI-based correction

Education

• MALLA REDDY COLLEGE OF ENGINEERING(MRCE)

  Bachelor's degree, Computer Science

• Kennesaw State University

  Master of Science - MS, Computer Science