About

9 years of hands-on expertise in AI/ML solution design and deployment. I specialize in architecting Deep Learning, Agentic AI, Large Language Models (LLMs), and GraphRAG systems that deliver measurable impact in production environments.

Currently building innovation through scalable GenAI architectures and MultiAgent AI solutions at IRIS Softwares Inc. I bridge the gap between stakeholder vision and technical feasibility.

Key Achievements

NVIDIA GTC Speaker 2024 & 2025
Olympic Games Security Deployment
US DoD Recognition

Experience

GenAI Solutions Architect

May 2025 - Present

IRIS Softwares Inc.

Designing cutting-edge GenAI solutions using Microsoft Azure and AWS. Building Multiagent AI systems with Langgraph, Autogen CrewAI, ACP, MCP and A2A.

Lead MLE (AI Centre of Excellence)

May 2021 - May 2025

Deloitte Offices of US in India

Built and deployed 4 LLM, RAG, VectorDB, GenAI-powered solutions for threat analysis. Developed real-time Cyber GenAI solutions with NVIDIA Morpheus, NIMs, NeMO. Presented Cyber AI solutions to US DoD and at NVIDIA GTC conferences.

Senior Data Scientist

Apr 2019 - Apr 2021

Sopra Steria India

Improved pipeline orchestration efficiency by 25% through AutoML optimization. Led research in NLP and pipeline orchestration technologies. Interfaced with AIRBUS for product delivery timeline optimization.

Senior Member of Technical Staff

Oct 2016 - Mar 2019

NEC Technologies India

Led a team of 5 in AI solution design and deployment. Converted client requirements into tailored AI solutions using Python and Deep Learning. Applied NLP, Word embeddings and Similarity analysis for complex problem solving.

Technical Skills

Generative AI

LLM RAG GraphRAG LightRAG Text-to-SQL Agentic AI Prompt Engineering

GenAI Tools

AWS Bedrock LangChain OpenAI APIs NVIDIA NIMs NeMO

Databases

MongoDB AWS Athena GCP BigQuery PostgreSQL Qdrant Milvus

NLP & ML

Chatbot NER BERT Transformers LLMs

Accelerated AI/ML

Rapids cuDF cuML NVIDIA TensorRT Dask-cuDF Dask Apache Spark

Graphs

GNN Neo4J PyTorch Geometric NetworkX

ML Frameworks

TensorFlow PyTorch XGBoost Keras MLFlow Pandas Scikit-Learn

Cloud Platforms

AWS Azure GCP DataBricks Lambda SageMaker Glue Athena Redshift Bedrock

Model Deployment

TF Serving NVIDIA Triton AWS Inferentia Flask API

Containerization

Docker GitHub Actions Kubernetes Kubeflow NVIDIA Morpheus Airflow

Monitoring

MLFlow Robust Intelligence

Featured Projects

LLM Powered Vulnerability Management

+20% faster resolution

RAG-powered solution for vulnerability prioritization with interactive chatbot interface.

AWS Bedrock OpenSearch RAG

Graph RAG Policy Assessment

+25% accuracy

LightRAG pipeline for policy document assessment using Claude 3.5, Milvus, and Neo4J.

LightRAG Neo4J Claude 3.5

Real-time Threat Detection

+70% faster triage

AI-enabled lateral movement detection using graph-based behavior modeling.

Graph ML NVIDIA Morpheus Analytics

Zero-Day Threat Detection

+20% throughput

Advanced encoder-decoder architectures with graph features for zero-day detection.

Deep Learning Graph Features NVIDIA

Airbus Delivery Prediction

95% accuracy

Deep learning solution for spare parts delivery prediction with FastAPI deployment.

Deep Learning FastAPI Prediction

RL for Penetration Testing

+20% citations

Reinforcement learning approach for exposing surveillance detection routes.

RL Attack Graphs Cyber Terrain

Threat Report Generation & Summarization

5x faster processing

Building real-time threat alert prioritization, summarization and enrichment using NVIDIA LLMs and RAG pipelines.

NVIDIA LLMs RAG Real-time

LLM Chatbot for Client Onboarding

50% reduced time

Built a chatbot using Microsoft Bot Framework, RASA and LLMs to enable real time responses for workshop onboarding experience.

Microsoft Bot RASA LLMs

AI Requirements Doc Generation

80% time saved

AI assisted Auto-generation of requirements and test cases to accelerate app integration/onboarding.

AutoML Requirements Test Cases

Custom NER for Invoices

30% efficiency gain

Enhanced invoice processing efficiency through custom CNN-CRF entity extraction techniques and improved entity prediction accuracy by 25%.

CNN-CRF NER Invoice Processing

Image Denoising for Airbus

70% noise reduction

Increased sharpness of image, removed unwanted lines and removed blurriness in invoice images by 70% using Auto-encoders.

Auto-encoders Image Processing Denoising

Publications & Research

Discovering exfiltration paths using reinforcement learning with attack graphs

December 2022

IEEE Conference on Dependable and Secure Computing (DSC)

View Paper

Exposing surveillance detection routes via reinforcement learning, attack graphs, and cyber terrain

November 2022

21st IEEE International Conference on Machine Learning and Applications (ICMLA)

View Paper

Lateral Movement Detection Using User Behavioral Analytics

July 2022

arXiv preprint arXiv:2208.13524

View Paper

Zero-Day Threat Detection Using Graph and Flow Based Security Telemetry

2022

International Conference on Computing, Communication, and Intelligent Systems (ICCCIS)

View Paper

Cross-Platform Lateral Movement Detection via Unsupervised Graph Machine Learning

2025

International Symposium on Digital Forensics and Security (ISDFS) 2025

View Paper