Generative AI ArchitectLLMs • Agents • RAG • MLOps

I design and build production-grade generative AI systems with strong observability and robust system architectures.

From multi-agent LLM orchestration systems to enterprise RAG platforms and cloud-native MLOps pipelines, I design and deploy production generative AI systems for real-world applications.

How I approach AI architecture, from design to deployment.

Agentic Systems

Multi agent orchestration, tool use, structured outputs, and guardrails.

Grounded GenAI

RAG pipelines, evidence grounding, evaluation, and hallucination mitigation.

Cloud & MLOps

Deployments on Azure/AWS with CI/CD, monitoring, and compliance first design.

About

I am an Applied AI Engineer at Deloitte, where I work on cloud-based multi-agent large language model (LLM) systems for federal healthcare and enterprise clients. My work focuses on designing and deploying generative AI systems that operate in real production environments.

In practice, this involves designing agentic workflows using LangGraph and LangChain, integrating retrieval-augmented generation (RAG), and developing end-to-end AI systems from model development to production deployment across cloud platforms.

Alongside industry work, I have a strong academic foundation in deep learning and applied machine learning. I have published peer-reviewed research with Springer on medical image classification using fine-tuned deep learning models and have hands-on experience deploying machine learning systems beyond research prototypes.

I am also the founder of the Data Science Club at Charusat University, where I led initiatives in applied machine learning, collaborative learning, and hands-on technical projects. These experiences shaped how I approach AI today: as an engineering discipline grounded in theory, systems thinking, and real-world constraints.

This portfolio highlights selected projects, architectures, and research that showcase how I design, implement, and reason about production-grade AI systems.

Founder, Data Science Club — Charusat University · Club website

Experience

Building production-grade AI systems at scale.

Generative AI Developer

Deloitte Consulting · Pennsylvania, United States

Oct 2023 – Present

Leading the design and deployment of production-grade generative AI systems for federal healthcare clients, focusing on reliability, scalability, and real-world impact.

  • Designed and deployed multi-agent LLM orchestration systems for NIH, CDC, and VBA.
  • Delivered AI-driven product development supporting new CDC contracts.
  • Built monitoring, logging, and alerting for production AI services.
  • Reduced service latency by 76% through async workflows and token optimization.
  • Owned CI/CD pipelines and operational readiness.
  • Implemented LLMOps / MLOps pipelines with evaluation and retraining hooks.
  • Enhanced distributed LLM workflows using MCP.
LLMsAgentsRAGLLMOpsMLOpsAWSAzureDistributed Systems

Data Analyst

Eminence Technology Solutions · New Jersey, United States

Dec 2022 – Jun 2023

Focused on building and optimizing large-scale data pipelines supporting AI and ML workflows in healthcare.

  • Built pipelines processing ~6 TB of healthcare data.
  • Automated ETL workflows across Spark clusters.
  • Streamlined data pipelines for faster model development.
Data PipelinesSparkETLCloudHealthcare Data

Machine Learning Intern

Fractal Analytics · New York, United States

Jun 2022 – Dec 2022

Worked on deploying and optimizing production ML models with monitoring and evaluation.

  • Deployed production ML models with monitoring.
  • Built a graph-based recommendation system.
Machine LearningModel MonitoringGraph Systems

Education

Academic foundation supporting applied AI engineering and large-scale data systems.

Bachelor of Technology in Information Technology

Charusat University · Gujarat, India

2019 – 2023

Undergraduate program focused on computer science fundamentals, including software engineering, algorithms, and applied information technology.

Computer ScienceSoftware EngineeringData Structures

Projects

Grounded Research Assistant

An agentic, evidence grounded research assistant that converts research papers into structured, traceable outputs using RAG.

  • Agentic workflow (plan → fan out → review → compile)
  • Evidence grounded outputs with retrieval
  • Interactive architecture diagram + HITL loop
Parallel AgentsRAGLangGraphHITL

Autonomous Debugger Assistant

A deterministic, multi-agent AI system that autonomously analyzes software failures, identifies root causes, generates safe code fixes, and validates them using a LangGraph-controlled execution loop.

  • Evaluator-driven bounded retry loop ensuring convergence
  • Structured system for coordinating multi-agent workflow
  • Safety guardrails for controlled autonomous code changes
LangGraphGuardrailsLLMOpsSystem Design

BoneX — Multi Bone Fracture Detection System

An AI system for detecting and classifying bone fractures across seven anatomical regions from X ray images, deployed as a usable web application.

  • 7 fracture categories supported
  • 84% precision/recall (vs 79% baseline)
  • Deployed workflow: upload → diagnosis
Computer VisionMedical ImagingTransformers

TravelBuddy — Multi Agent AI Travel Assistant

A multi agent AI travel planning system that routes user requests to specialized agents for efficient, low latency, and cost aware task execution.

  • Supervisor agent orchestration
  • Intent based routing (LLM vs tools)
  • Latency and cost optimized execution
LLM OrchestrationCost OptimizationLangchain

Publications

Classification of Choroidal Neovascularization (CNV) from Optical Coherence Tomography (OCT) Images Using Efficient Fine-Tuned ResNet and DenseNet Deep Learning Models

Age-related macular degeneration (AMD) is a macular degenerative disease and a leading cause of blindness worldwide, often associated with choroidal neovascularization (CNV). This study focuses on the classification of CNV from Optical Coherence Tomography (OCT) images using efficient fine-tuned ResNet and DenseNet deep learning models. The proposed approach aims to accurately identify CNV to support early diagnosis and treatment. Model performance is evaluated and compared to determine the most effective architecture for CNV classification, contributing toward more reliable and efficient AI-assisted diagnostic tools in ophthalmology.

Venue: Springer — Intelligent Computing / Artificial Intelligence (Book Chapter)

Agent-Based Research Workflows for Evidence-Grounded Analysis

This technical report explores agent-based orchestration for automated research analysis, emphasizing evidence grounding, structured outputs, and iterative refinement. The workflow is designed to support reproducibility, critical review, and human-in-the-loop validation in research and applied AI systems.

Status: Technical report / preprint

Let’s collaborate on building intelligent solutions that matter.

Interested in AI systems and scalable architectures. Let’s connect.