Naga Sai Purvaz — AI/ML Engineer

AI/ML Engineer · Generative & Agentic AI

 ███╗   ██╗ █████╗  ██████╗  █████╗     ███████╗ █████╗ ██╗
 ████╗  ██║██╔══██╗██╔════╝ ██╔══██╗    ██╔════╝██╔══██╗██║
 ██╔██╗ ██║███████║██║  ███╗███████║    ███████╗███████║██║
 ██║╚██╗██║██╔══██║██║   ██║██╔══██║    ╚════██║██╔══██║██║
 ██║ ╚████║██║  ██║╚██████╔╝██║  ██║    ███████║██║  ██║██║
 ╚═╝  ╚═══╝╚═╝  ╚═╝ ╚═════╝ ╚═╝  ╚═╝    ╚══════╝╚═╝  ╚═╝╚═╝

 ██████╗ ██╗   ██╗██████╗ ██╗   ██╗ █████╗ ███████╗
 ██╔══██╗██║   ██║██╔══██╗██║   ██║██╔══██╗╚════██║
 ██████╔╝██║   ██║██████╔╝██║   ██║███████║    ██╔╝
 ██╔═══╝ ██║   ██║██╔══██╗╚██╗ ██╔╝██╔══██║   ██╔╝
 ██║     ╚██████╔╝██║  ██║ ╚████╔╝ ██║  ██║   ██║
 ╚═╝      ╚═════╝ ╚═╝  ╚═╝  ╚═══╝  ╚═╝  ╚═╝   ╚═╝

Building autonomous agents · LLM pipelines · RAG architectures

3+ years · New York · Enterprise AI · 40%+ efficiency gains

$ npx naga-sai --role="AI/ML Engineer" --location="New York" --status="open"

✔ Loading profile · done in 3ms

Python

PyTorch

LangChain

AWS

Kubernetes

GPT-4/Claude

scroll ↓

Live · Jan 2025 – Present

CVS Health · Dallas, TX

Enterprise Agentic
AI Platform

AI/ML Engineer · Generative AI & Agentic Systems

Designing and shipping a full-scale production agentic AI platform — autonomous multi-step decision-making, stateful workflows, multi-LLM orchestration, and enterprise RAG serving real clinical and operational use cases.

Agentic pipelines: LangGraph, CrewAI, AutoGen, Agno, DSPy, Semantic Kernel — stateful execution, memory management, human-in-the-loop approvals.
Multi-LLM orchestration across GPT-4.1/4o, Claude, Gemini 1.5, LLaMA, T5 — dynamic cost/latency/task-aware routing.
Production RAG: FAISS, Pinecone, Weaviate, Neo4j knowledge graphs — hybrid dense + BM25, LangSmith observability, RAGAS/DeepEval metrics.
Fine-tuning: PEFT/LoRA/QLoRA domain adaptation, RLHF/DPO alignment and safety.
LLMOps: canary/blue-green CI/CD, vLLM + TensorRT-LLM + NVIDIA Triton, speculative decoding, intelligent caching.
Cloud-native: AWS SageMaker, Bedrock, EKS, EMR · Terraform 1.6+ / AWS CDK v2 · Airflow, Spark 3.5+, Kafka, NiFi, Databricks.

LangGraphCrewAI GPT-4oClaude LoRA/QLoRARLHF PineconeNeo4j BedrockSageMaker KafkaTerraform vLLMTriton

40%+

efficiency gain

agent frameworks

LLM models

cloud providers

Agent Modules

Multi-step clinical decision reasoning
Stateful task decomposition engine
Human-in-the-loop approval layer
Memory & context management
Dynamic LLM router (cost/latency)

Infra & Ops

vLLM + TensorRT-LLM inference optimization
NVIDIA Triton Inference Server
Canary & blue/green deployments (Jenkins)
Prometheus + Grafana + OpenTelemetry
Airflow + Spark + Kafka pipelines

Governance & Safety

Agent safety guardrails & RBAC
Red-teaming workflows
LLM token/cost/drift monitoring
Ethical AI filters & audit logging

02 Full Technology Arsenal

◈ GenAI & LLMs 12 models ›

GPT-4/4oGPT-3.5 Claude (Anthropic)Gemini LLaMA 3.1LLaMA 2 MistralGemma Phi-3Grok T5Ollama

◈ Agentic & GenAI Frameworks 10 frameworks ›

LangChainLangGraph LlamaIndexHugging Face Transformers Semantic KernelAutoGen CrewAIDSPy HaystackAgno

◈ RAG & Vector Databases 8 systems ›

PineconeFAISS WeaviateQdrant ChromaMilvus ElasticsearchNeo4j

◈ ML & Deep Learning 13 tools ›

PyTorchTensorFlow KerasScikit-learn LightGBMPEFT / LoRA / QLoRA RLHF / DPOvLLM TensorRT-LLMNVIDIA Triton ProphetARIMA DNN / CNN / RNN / LSTM

◈ Cloud Platforms 14 services · 3 clouds ›

AWS SageMakerAWS Bedrock AWS LambdaAWS S3 AWS EKSAWS EMR AWS RedshiftAWS Step Functions GCP Vertex AIGCP BigQuery GCP Pub/SubAzure OpenAI Azure MLAzure Synapse

◈ MLOps & Deployment 12 tools ›

MLflowDocker KubernetesFastAPI KubeflowRay Weights & BiasesBentoML KServeTorchServe ONNXFlask

◈ Data & Streaming 11 systems ›

Apache KafkaApache Spark 3.5+ Apache AirflowApache NiFi DatabricksRedis PostgreSQLMongoDB MySQLSnowflake Redshift

◈ Monitoring, Security & Observability 9 tools ›

PrometheusGrafana LangSmithCloudWatch OpenTelemetryRAGAS DeepEvalIAM / Okta Audit Logging

◈ DevOps & Infrastructure 7 tools ›

Terraform 1.6+AWS CDK v2 JenkinsGitHub Actions Blue/Green DeployCanary Deploy Infrastructure as Code

◈ Languages 9 languages ›

PythonSQL JavaJavaScript TypeScriptBash / Shell RHTMLXML

◈ NLP, CV & Visualization multi-domain ›

Sentiment AnalysisNER OCRSemantic Search EmbeddingsObject Detection Image ClassificationTime-Series Forecasting Anomaly Detection Power BITableau PlotlyMatplotlib

03 Projects

01 Agentic AI

Clinical Multi-Agent Decision System

Autonomous clinical NLP reasoning agents across medical records, lab data, and treatment protocols with human-in-the-loop checkpoints at CVS Health.

LangGraphGPT-4o BedrockFastAPI

02 RAG

Enterprise Knowledge Intelligence Platform

Hybrid dense + BM25 RAG with Neo4j knowledge graphs and LangSmith observability. Achieved 40%+ improvement in enterprise knowledge retrieval accuracy.

LlamaIndexNeo4j WeaviateRAGASFAISS

03 Fine-Tuning

Capital Markets Risk Intelligence LLM

Domain-adapted LLM for equities risk modeling via LoRA/QLoRA fine-tuning and RLHF alignment. Ultra-low latency inference deployed on NVIDIA Triton Inference Server.

LoRA/QLoRARLHF vLLMSageMakerTriton

04 Multi-LLM

LLM Orchestration Router

Intelligent routing platform dynamically selecting between GPT-4, Claude, Gemini, and open-source models based on cost, latency, and task context — with full LangSmith observability.

GPT-4.1Claude Gemini 1.5LangSmithOpenTelemetry

05 MLOps

Real-Time AI Pipeline Orchestration

Event-driven AI inference pipeline processing high-volume data streams with Apache Kafka and Spark 3.5+, orchestrated via Airflow on EKS with drift monitoring and canary deployments.

KafkaSpark 3.5+ AirflowKubernetesMLflow

06 Governance

Responsible AI Governance Framework

End-to-end AI governance system with RBAC, ethical AI filters, red-teaming workflows, agent safety guardrails, and real-time monitoring of LLM token usage, cost, and drift.

PrometheusGrafana IAM / OktaDeepEvalCloudWatch

04 Experience

Jan 2025 — Present

CVS Health

AI/ML Engineer · Generative AI & Agentic Systems · Dallas, TX

Enterprise Agentic AI platforms using LangGraph, CrewAI, AutoGen, Agno, DSPy — stateful execution, memory management, human-in-the-loop.
Multi-LLM orchestration: GPT-4.1/4o, Claude, Gemini 1.5, LLaMA — dynamic cost/latency routing.
Production RAG with FAISS, Pinecone, Weaviate, Neo4j — hybrid dense + BM25, RAGAS/DeepEval metrics.
LLMOps: canary/blue-green CI/CD, vLLM + TensorRT-LLM + NVIDIA Triton inference optimization.
AWS cloud-native infra (SageMaker, Bedrock, EKS, EMR) · Terraform 1.6+ / CDK v2 · Airflow, Spark 3.5+, Kafka, Databricks.

May 2022 — Nov 2023

NTT DATA

MLOps Engineer · Hyderabad, India

Cloud-native AI infra with Terraform and Kubernetes (EKS/AKS) for distributed training and multi-tenant workloads.
LLMOps: RAG pipelines, vector stores, PEFT/LoRA fine-tuning at scale.
Full ML lifecycle with MLflow — experiment tracking, registry, versioning, governance for PyTorch and TensorFlow.
Inference optimization: Docker, FastAPI, KServe, ONNX, TorchServe, caching strategies.

Nov 2019 — May 2020

HSBC

DevOps Engineer · Hyderabad, India

AWS infra (EC2, Lambda, S3, VPC) with Terraform IaC. CI/CD via Jenkins + GitHub Actions.
Docker + Kubernetes Blue/Green and Rolling deployments. CloudWatch + Grafana monitoring. IAM security.