Arpit Agarwal - AI Research Engineer & Patent Inventor

About Arpit Agarwal

AI Engineer & Data Scientist with 6+ years of experience in MCP, LLMs, LangChain, LangGraph, RAG, agentic AI, fine-tuning and LangSmith. Specialized in building AI agents, autonomous workflows, and enterprise-grade ML pipelines. Experienced in DBs, Kafka, Apigee, AWS, and Kubernetes, delivering scalable, real-time AI solutions that cut costs, accelerate deployments, and drive measurable impact.

💰 $5M+ Cost Savings: Delivered through AI automation and optimization

🚀 50% MTTR Reduction: Built "Booster" AI agent federating 13 databases

⚡ 40,000+ Hours Saved: Automated 5G site deployment validation

📊 1.5M Customer Migration: Led with near-zero downtime

🛠️ Technology Stack

Agentic AI & Multi-Agent Systems

CrewAI, AutoGPT, LangGraph workflows

LLM Fine-tuning & Optimization

LoRA, QLoRA, PEFT, Unsloth, vLLM

Vector & Graph Databases

ChromaDB, Pinecone, Neo4j, Qdrant, FAISS

🏆 Patents Portfolio

AI-Assisted Adjustment of a 5G Network

US20250081042A1

Revolutionary AI system for optimizing 5G network performance

Impact: Advanced telecommunications infrastructure

AI/ML 5G Networks Network Optimization

Intelligent Synthetic Data Injection

US20250081006A1

Innovative approach to synthetic data generation and injection

Impact: Enhanced AI model training and data privacy

Synthetic Data AI/ML Data Generation

Extreme Validation for Fault Detection

US20250080395A1

Advanced fault detection system with extreme validation methods

Impact: Improved system reliability and error detection

Fault Detection Validation Systems AI/ML

Key Projects

Agentic AI Framework

Advanced multi-agent AI system with autonomous reasoning

Python LangChain Multi-Agent

View Repository

LangGraph Solution Generator

Intelligent solution generation using LangGraph workflows

Python LangGraph Workflows

View Repository

Summarize and Validator

AI-powered content summarization and validation system

Python NLP Validation

View Repository

ML Project V1

Comprehensive machine learning project with end-to-end pipeline

Python Machine Learning MLOps

View Repository

Skeeball Game

Interactive web-based skeeball game with physics simulation

JavaScript HTML5 Canvas Game Dev

Play Game View Repository

Golden Map Game

Adventure treasure hunting game with interactive map exploration

JavaScript HTML5 Adventure

Play Game View Repository

Testing as a Service

Revolutionary approach to testing - not just to validate, but to learn and improve systems continuously

Testing Strategy DevOps Quality Assurance

Read Article

💼 Professional Experience

Jul 2025 - Present

Senior AI Engineer

EchoStar Corporation, Denver, CO

Booster AI Agent (LLM | MCP | Agentic AI | LangGraph | LangSmith)

Built "Booster" Supervised agentic AI assistant using MCP and LangGraph, federating 13 databases and reducing MTTR & truck rolls by 50%
Developed predictive dispatch & route-optimization system that scored at-risk locations, optimized engineer routing, and enabled real-time rerouting — $1M+ annual OPEX savings
Scaled microservices (Docker + OpenShift + CI/CD) → achieved 3× throughput and reduced p95 latency

Customer Experience Chatbot (Mistral | Fine-tuning | LangServer | Bedrock)

Deployed CXO-facing chatbot using fine-tuned (Quantized) Mistral LLM with persistent memory and human-in-loop reinforcement
Reduced order fallout resolution time from 60 minutes to 2 minutes while improving user satisfaction and accuracy
Saved 5M tokens/day and reduced operational costs by building intelligent scheduling pipeline

Feb 2023 - Jul 2025

Data Scientist II

Boost Mobile, Riverfront, CO

Site Deployment (Lambda | State Machine | DynamoDB | Deep Learning)

Automated 5G site deployment validation with embedding + transformer models for anomaly detection and rule-based validation → eliminated manual effort, saving 40,000+ hours
Led 1.5M customer network migration (Lambda, Step Functions, Kafka, Apigee APIs) → delivered $100k savings with near-zero downtime

Incident Management (WatsonX | React | APIs | Milvus | CI/CD | Bedrock)

Developed incident management system (WatsonX + RAG) mapping child/parent tickets; achieved 80% accuracy and reduced duplicate work
Leveraged GenAI and ML clustering to group orphan incident tickets with similar features and automatically assign/create parent tickets, saving 1,400 hours of manual work weekly

Supply Chain API Insights (Spark | AWS | Glue | Redshift | NLP)

Scaled supply chain ETL pipelines (Spark + AWS Glue) to handle 10k+ requests/min, enabling real-time analytics and faster decision-making
Automated ServiceNow scheduling workflows (calendar creation, rescheduling, cancellations), enabling scalable vendor onboarding and saving $1.5M annually

Sep 2021 - Feb 2023

R&D - Data Scientist II

Dish Wireless, Littleton, CO

Testing as a Service (AWS | Data Platform | Neural Network | SageMaker)

Architected Testing-as-a-Service (TaaS) platform validating CNFs via AWS Kinesis → DynamoDB → SageMaker pipeline, cutting resolution time by 10 days and saving $3M/year
Developed neural network models injecting synthetic loads to simulate 1 Billion users, identifying CNF breakpoints for optimized capacity planning

May 2018 - Jul 2019

Data Analyst

Mahindra & Mahindra, India

Built predictive models to forecast delivery readiness and detect potential quality issues across 22,500+ tractors and 30+ inspection points
Developed interactive dashboards (Python/SQL/Excel/Tableau) to optimize procurement and production planning, supporting 57% YoY sales growth

Jul 2017 - Jul 2019

CEO & Founder

Tafri, Jaipur, India

Launched travel itinerary & bucket-list platform with 100+ active subscribers, featuring a recommendation engine with intelligent reminders

📝 Publications & Media

"Testing as a Service: Test Not Only to Validate but to Learn"

DISH Developer Experience

Read Article

"Wall Street is Fed Up with Dish Network"

Featured in Light Reading

Media Mention: Technical contributions recognized

Technical Expertise

AI & Machine Learning

Agentic AI Systems (CrewAI, AutoGPT)
LLM Fine-tuning (LoRA, QLoRA, PEFT)
Multi-Context Protocol (MCP)
RAG Systems & Vector Search
Synthetic Data Generation
Small Language Models (SLM)
Computer Vision (OpenCV, YOLO)
Natural Language Processing

Data Engineering & Databases

Vector Databases (ChromaDB, Pinecone, Qdrant)
Graph Databases (Neo4j, ArangoDB)
Time Series (InfluxDB, TimescaleDB)
Big Data (Spark, Kafka, Airflow)
Data Pipelines & ETL
Real-time Analytics

MLOps & Cloud Platforms

MLflow, Kubeflow, DVC
Docker, Kubernetes
AWS (SageMaker, Bedrock, Lambda)
GCP (Vertex AI, BigQuery)
Model Deployment & Monitoring
CI/CD for ML Pipelines

Programming & Frameworks

Python (PyTorch, TensorFlow, Transformers)
LangChain, LlamaIndex, Haystack
FastAPI, Streamlit, Gradio
JavaScript, React, Node.js
SQL, NoSQL, GraphQL
Git, Linux, Shell Scripting

Research & Innovation

Patent Development & Filing
AI Research & Publications
Experimental Design
Statistical Analysis
A/B Testing & Validation
Technical Writing

📊 GitHub Analytics

🏆 Awards & Recognition

🥇

GenAI Q2 2025

EchoStar Corporation

🎮

Agentic AI Gaming

Amazon Recognition

🏆

Hackathon Winner

Boost Mobile

👨‍💼

Technical Leadership Award

Dish Wireless

💰

Leadership Award

Dish Network Finance

🔄

Winning Award Migration

OSS Systems

🔒

Security Vulnerability Discovery

Critical SMTP Security Issue

📜

3 US Patents

AI & 5G Network Innovation

💼 Professional Services

Agentic AI Development

Custom autonomous AI agents and multi-agent systems

LLM Fine-tuning Services

Domain-specific model customization with LoRA/QLoRA

Enterprise RAG Solutions

Advanced retrieval systems with vector databases

MCP Implementation

Multi-context protocol for complex AI workflows

Get in Touch

Let's collaborate on innovative AI solutions and breakthrough technologies.

GitHub Email LinkedIn