Viswajit Kudugunti

About Viswajit Kudugunti

Applied AI/ML Engineer with 4+ years of experience building and deploying large-scale LLM and Retrieval-Augmented Generation (RAG) systems across organizations like Meta, Perplexity, and Accenture.I specialize in developing production-grade GenAI applications using LLaMA 3.x, LangChain, and AWS, with a strong focus on scalable AI infrastructure, real-time inference, and enterprise search systems.My work includes: Fine-tuning LLMs using QLoRA, DeepSpeed, and FSDP Designing high-performance inference pipelines with sub-200ms latency Building scalable RAG systems across large enterprise datasets Developing multi-agent AI workflows and LLM APIs Implementing MLOps pipelines for faster model deployment cyclesI have hands-on experience across the full ML lifecycle—from model training and optimization to deployment, monitoring, and scaling in production environments.Passionate about building reliable, high-performance GenAI systems that combine contextual intelligence, personalization, and real-world impact.

Experience

1. Ai Ml Engineer

   Perplexity

   Jan 2025 — Present

   Worked on large-scale LLM systems and RAG architectures for enterprise AI search, focusing on performance optimization, scalability, and reliability.Fine-tuned LLaMA 3.x models using QLoRA, DeepSpeed, and FSDP, improving domain QA accuracy by 37% Built high-performance inference pipelines achieving 1100+ tokens/sec and <180ms latency| Designed scalable RAG systems across 10B+ documents, improving factual accuracy to 93% Reduced hallucinations by 18% using RLAIF and prompting strategies Developed multi-agent pipelines improving user engagement by 27% Engineered AWS-based AI infrastructure with 99.5% uptime Accelerated model release cycles by 35% via MLOps pipelines

Education

• Webster University

  Master's degree, Data Analytics