NAGA SAI LILITH JEJI BABU KARRI

Achieved a 25% reduction in infrastructure spend by implementing serverless LLM APIs on AWS Lambda and SageMaker.

Enhanced model throughput by optimizing transformers with prompt caching and PyTorch Lightning, improving processing efficiency.

Developed and deployed LSTM-based anomaly detectors using AutoGen, decreasing false positives by 45%.

Engineered high-accuracy churn prediction models (XGBoost) with 82% accuracy, contributing to an 18% reduction in system downtime.

Developed and deployed a multi-agent RAG platform (CrewAI, LangChain), cutting repeat customer tickets by 35% and significantly enhancing overall customer satisfaction.

Optimized large language models by fine-tuning Llama-2 (LoRA/QLoRA) with adaptive RAG, resulting in a 40% reduction in AI hallucination rates.

Architected and delivered a robust voice-enabled AI system utilizing LangGraph, achieving 99.2% uptime and supporting over 10 million monthly interactions.

Reduced cloud inference costs by 35% through strategic migration to AWS Inferentia and deployment of highly efficient serverless LLM APIs.

Deployed advanced computer vision safety models (OpenCV, TensorFlow), leading to a 22% reduction in workplace incidents.

Constructed an IoT predictive maintenance pipeline using Spark and Python, achieving 78% accuracy in forecasting equipment failures.

Pioneered the implementation of edge AI on Jetson Nano and Raspberry Pi, enabling low-latency IoT analytics for real-time insights.

Developed robust ML pipelines using scikit-learn and Python for real-time workforce safety monitoring.

Reduced model training time by 25% through the application of advanced feature engineering techniques.

Implemented unsupervised anomaly detection and NLP compliance workflows, enhancing operational efficiency by 20%.

Prototyped machine learning models with TensorFlow and PyTorch, accelerating deployment cycles by 30%.