Looking for a job

Priyayarramreddyai User

@priyayarramreddyaius

AI/ML Engineer at AMD building enterprise RAG and agentic AI systems, including a platform processing 2M+ daily queries.

United States

Message

At AMD, I architected an enterprise RAG platform that processes 2M+ daily queries and reduced search latency from 1.2 seconds to 180 milliseconds. I also engineered agentic AI applications and optimized GPU-accelerated LLM inference, increasing GPU utilization from 60% to 94%.

At Accenture, I led development of a customer churn platform for a telecom client serving 2.5M subscribers, achieving an AUC of 0.89 and generating USD 8.2M in annual business value. Earlier, at Cognizant, I developed computer vision solutions that reached 94.2% detection accuracy across eight manufacturing lines.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Engineer

Current

AMD

May 2025 - Present (1 year 6 months)

Architected an enterprise RAG platform using Claude API, LangChain, FAISS, and Hybrid Search, processing 2M+ daily queries. Led LLM fine-tuning with QLoRA and PEFT, and optimized GPU inference with CUDA and vLLM.

RAG Lang Chain Faiss Hybrid Search Embeddings QLoRA PEFT LLaMA A100 ROCm CUDA VLLM LangGraph LlamaIndex MCP Docker Kubernetes MLFlow GitHub Actions Prometheus Grafana LLM As A Judge Prompt Engineering

AI Engineer

Accenture

May 2020 - Aug 2023 (3 years 3 months)

Led development of a real-time Customer Churn Prediction platform for a telecom client, achieving an AUC of 0.89. Built scalable feature engineering pipelines with PySpark and Databricks, and developed demand forecasting models using LSTM networks.

XGBoost PySpark Databricks LSTM BERT NLP fastAPI PostgreSQL REST APIs Feature Engineering Machine Learning

AI Software Engineer

Cognizant

Jul 2018 - Apr 2020 (1 year 9 months)

Developed computer vision solutions using YOLOv3 and OpenCV, achieving 94.2% detection accuracy. Built demand forecasting models with stacked LSTM networks and an anomaly detection platform using Isolation Forest and LSTM autoencoders.