Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I'm Durga Prasad, a Senior AI/ML Engineer and GenAI Architect with 8+ years of experience designing, building, and deploying production-grade ML systems, Generative AI applications, and LLMOps infrastructure across enterprise, healthcare, and financial domains. I specialize in LLM fine-tuning (LoRA, QLoRA, DPO, RLHF), Agentic multi-agent systems, RAG and GraphRAG architectures, and scalable real-time ML pipelines using Kafka, PySpark, and Flink. I’ve delivered measurable improvements in model accuracy, latency, and cost, while upholding strict security and compliance standards. My work spans across government, healthcare, and financial sectors with hands-on expertise in AWS Bedrock, Azure AI Foundry, and GCP Vertex AI, plus advanced inference optimization tools like vLLM, TensorRT-LLM, and Triton. I’m passionate about building robust, observable AI systems and enabling data-driven decision making at scale.…I'm Durga Prasad, a Senior AI/ML Engineer and GenAI Architect with 8+ years of experience designing, building, and deploying production-grade ML systems, Generative AI applications, and LLMOps infrastructure across enterprise, healthcare, and financial domains. I specialize in LLM fine-tuning (LoRA, QLoRA, DPO, RLHF), Agentic multi-agent systems, RAG and GraphRAG architectures, and scalable real-time ML pipelines using Kafka, PySpark, and Flink. I’ve delivered measurable improvements in model accuracy, latency, and cost, while upholding strict security and compliance standards. My work spans across government, healthcare, and financial sectors with hands-on expertise in AWS Bedrock, Azure AI Foundry, and GCP Vertex AI, plus advanced inference optimization tools like vLLM, TensorRT-LLM, and Triton. I’m passionate about building robust, observable AI systems and enabling data-driven decision making at scale.

DURGA PRASAD d.nelakurthi

AI Engineer, Data Scientist, Data Analyst, +2





I'm Durga Prasad, a Senior AI/ML Engineer and GenAI Architect with 8+ years of experience designing, building, and deploying production-grade ML systems, Generative AI applications, and LLMOps infrastructure across enterprise, healthcare, and financial domains. I specialize in LLM fine-tuning (LoRA, QLoRA, DPO, RLHF), Agentic multi-agent systems, RAG and GraphRAG architectures, and scalable real-time ML pipelines using Kafka, PySpark, and Flink. I’ve delivered measurable improvements in model accuracy, latency, and cost, while upholding strict security and compliance standards. My work spans across government, healthcare, and financial sectors with hands-on expertise in AWS Bedrock, Azure AI Foundry, and GCP Vertex AI, plus advanced inference optimization tools like vLLM, TensorRT-LLM, and Triton. I’m passionate about building robust, observable AI systems and enabling data-driven decision making at scale.…I'm Durga Prasad, a Senior AI/ML Engineer and GenAI Architect with 8+ years of experience designing, building, and deploying production-grade ML systems, Generative AI applications, and LLMOps infrastructure across enterprise, healthcare, and financial domains. I specialize in LLM fine-tuning (LoRA, QLoRA, DPO, RLHF), Agentic multi-agent systems, RAG and GraphRAG architectures, and scalable real-time ML pipelines using Kafka, PySpark, and Flink. I’ve delivered measurable improvements in model accuracy, latency, and cost, while upholding strict security and compliance standards. My work spans across government, healthcare, and financial sectors with hands-on expertise in AWS Bedrock, Azure AI Foundry, and GCP Vertex AI, plus advanced inference optimization tools like vLLM, TensorRT-LLM, and Triton. I’m passionate about building robust, observable AI systems and enabling data-driven decision making at scale.

Available to hire

I’m Durga Prasad, a Senior AI/ML Engineer and GenAI Architect with 8+ years of experience designing, building, and deploying production-grade ML systems, Generative AI applications, and LLMOps infrastructure across enterprise, healthcare, and financial domains. I specialize in LLM fine-tuning (LoRA, QLoRA, DPO, RLHF), Agentic multi-agent systems, RAG and GraphRAG architectures, and scalable real-time ML pipelines using Kafka, PySpark, and Flink.

I’ve delivered measurable improvements in model accuracy, latency, and cost, while upholding strict security and compliance standards. My work spans across government, healthcare, and financial sectors with hands-on expertise in AWS Bedrock, Azure AI Foundry, and GCP Vertex AI, plus advanced inference optimization tools like vLLM, TensorRT-LLM, and Triton. I’m passionate about building robust, observable AI systems and enabling data-driven decision making at scale.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Language

English

Fluent

Work Experience

Senior AI/ML Engineer & GenAI Architect at State of New York

November 1, 2025 - Present

Architected a production LangGraph + CrewAI multi-agent orchestration system using GPT-4o and Claude 3.7 Sonnet to automate RFP generation, compliance validation, and cross-agency knowledge synthesis; reduced analyst workload by 45% and cut document review cycle from 5 days to 8 hours. Designed a GraphRAG architecture with Neo4j, Qdrant, and Azure AI Search enabling multi-hop reasoning over 200GB+ of unstructured government documents with sub-second retrieval latency and 89%+ answer accuracy validated via RAGAS. Fine-tuned Llama 3.1-70B and Mistral-Large using QLoRA + DPO via Axolotl and Unsloth on domain-specific datasets, improving contextual accuracy by 38% and reducing hallucinations by 62% vs zero-shot. Built end-to-end LLM observability with Langfuse, Arize Phoenix, and Weights & Biases tracking cost, latency, prompt quality, and drift across 20+ production models, achieving sub-100ms P95 latency and 4x throughput with TensorRT-LLM optimizations. Implemented Guardrails AI and cus

Senior AI/ML Engineer at Cigna Healthcare

October 1, 2023 - September 30, 2025

Built HIPAA-compliant LLM-powered healthcare applications using GPT-4o, LangChain, and RAG pipelines for intelligent clinical document retrieval, medical summarization, and ICD-10 code suggestion — reducing clinical documentation effort by 40%. Fine-tuned BioBERT and Llama 3 on proprietary EHR datasets using LoRA + PEFT (TRL) to improve clinical NER F1-score by 22% and medical summarization ROUGE-L by 18%. Engineered AWS Kinesis + Kafka + Flink pipelines ingesting 2M+ daily EHR/claims events with Protobuf serialization, reducing data-to-insight latency from 6 hours to under 4 minutes. Architected end-to-end MLOps using SageMaker Pipelines, MLflow, Docker, and Kubernetes with RBAC, field-level encryption, and audit logging to achieve SOC 2 Type II and HIPAA compliance. Optimized transformer inference with ONNX Runtime (INT8/FP16) reducing average latency by 55% and GPU costs by ~$180K/year. Implemented Feast-based online/offline feature stores (150+ features across 8 models) and deliv

AI Data Scientist at Wells Fargo

January 1, 2022 - September 30, 2023

Built sub-200ms real-time credit risk scoring using XGBoost + LightGBM served via Ray Serve and FastAPI on AWS SageMaker, processing 500K+ daily loan applications with 0.93 AUC-ROC and SHAP-based explanations. Developed unsupervised anomaly detection (Isolation Forest, Autoencoder) for real-time fraud, reducing false positives by 28% and saving $12M+ annually. Implemented SageMaker Pipelines + MLflow automated retraining with statistical drift detection (PSI, KS) across 3+ production models, reducing degradation incidents by 70%. Engineered 200+ features from transactional, behavioral, and credit bureau data (PySpark/Databricks), improving credit default prediction precision by 18% across 5M+ portfolios. Used K-Means + UMAP to segment 8M+ customers into behavioral cohorts for targeted campaigns (24% improvement). Delivered risk analytics dashboards via Power BI/Tableau.

Data Scientist at Hexaware Technologies

December 1, 2017 - June 30, 2021

Engineered end-to-end ML pipelines using Scikit-learn and XGBoost for churn prediction and demand forecasting; deployed 10+ models as FastAPI microservices on Kubernetes for real-time predictions under ~150ms. Built Hadoop-based ETL pipelines processing 500GB+ daily, reducing data prep time by 60% and improving data quality to 99.5%+. Implemented an A/B testing framework for model champion/challenger evaluations, validating 20+ iterations with confidence intervals and business impact. Delivered Tableau/Power BI dashboards for KPI tracking and model performance, adopted by multiple enterprise clients; optimized Spark/Hadoop workflows to boost data throughput by 3x and cut latency by 50%.