tanay.

TANAYAGARWAL.

AI Engineer·Multi-Agent Systems·LLM Orchestration

I build intelligent AI systems that automate work and deliver real results. Currently crafting production-ready multi-agent ecosystems at TheAgentic.

Who I am

I'm an AI Engineer who builds production-ready intelligent systems that solve real-world problems. Currently at TheAgentic (Zenith Software Limited), I specialize in building multi-agent AI systems using PydanticAI, enabling autonomous document intelligence, structured data extraction, and agent-orchestrated reasoning.

I've developed enterprise-scale autonomous agent ecosystems (15+ agents) for CRM, onboarding, and lifecycle analytics—delivering end-to-end AI automation for business operations across platforms like HubSpot, Salesforce, Slack, Gmail, and Jira.

Focus

  • Multi-Agent AI Systems
  • Enterprise Automation
  • Data Engineering

Approach

  • Production-ready AI solutions
  • Strategic decision-making
  • Client-centric development

AI Experiments

  • Tinkering with new LLMs
  • Building personal AI agents
  • Exploring GPT-4 & Claude

Where I've worked

TheAgentic(Zenith Software Limited)

Artificial Intelligence Developer

May 2025 – Present

Built and productionized multi-agent AI systems using PydanticAI, enabling autonomous document intelligence, structured data extraction, and agent-orchestrated reasoning.

  • Developed enterprise-scale autonomous agent ecosystems (15+ agents) for CRM, onboarding, and lifecycle analytics—integrated with HubSpot, Salesforce, Slack, Gmail, Jira.
  • Designed advanced retrieval-augmented reasoning, policy-intelligence, and strategic-recommendation agents for insurance and financial domains.
  • Implemented multimodal AI pipelines, including vision-augmented agents and web-intelligence agents for deep research and risk analysis.
  • Coordinated directly with clients to refine requirements, optimize performance, and deliver production-ready AI solutions.
PydanticAIMulti-Agent AILLMsRAGEnterprise Automation

Saint Gobain – INDEC

Associate Software Engineer

Jun 2024 – May 2025

Acquired hands-on expertise in Data Engineering tools and designed Generative AI Chatbots for enterprise use.

  • Acquired expertise in Snowflake, Azure Data Factory (ADF), and Azure Synapse Analytics.
  • Designed three Generative AI Chatbots: HR Chatbot, BookOfIT Chatbot, and ServiceNow Chatbot.
  • Leveraged LangChain, LangGraph, and Knowledge Graphs for efficient, context-aware chatbots.
SnowflakeAzure ADFLangChainLangGraphKnowledge Graphs

Shunya Dot Ek

Machine Learning Intern

Feb – Aug 2023

Increased product accuracy to 97.6% using deep learning for computer vision.

  • Implemented deep learning models for product recognition, object tracking, and movement recognition.
  • Utilized YOLO, Faster-RCNN, and OpenCV for computer vision tasks.
Computer VisionYOLODeep LearningOpenCVPyTorch

ICICI Prudential

Software Engineer Intern

Aug – Dec 2022

Decreased Bloomberg file transfer time by over 40% through automation.

  • Automated project payment file transfer, reducing transfer time by 40%+.
  • Assisted in Host-to-Host File Transfer Automation for Investment Rationale Project.
SAPAutomationFile TransferBloomberg

What I'm working on

VoxAgent

In Active Development

Real-time AI voice telecalling system that runs a fully autonomous outbound sales call pipeline — live microphone capture, streaming speech-to-text, LLM-driven conversation, and instant text-to-speech playback — all with sub-second latency.

Voice Pipeline

1
Deepgram STT

Streaming speech-to-text with dual-signal turn detection and incomplete-utterance guards

2
Groq LLM

Speculative token streaming on Llama 3.3 70B with persona-driven call flow and objection handling

3
Cartesia TTS

Sonic-3 streaming text-to-speech with echo-aware VAD for natural barge-in support

Highlights

  • End-to-end voice loop: Mic → STT → LLM → TTS → Speaker in real-time
  • Speculative LLM inference — starts generating before STT finalizes for ~300ms head start
  • Echo-aware VAD enables barge-in without headphones
  • Structured call logging with delta timestamps for latency debugging
  • Full call recording and transcript export for review and tuning
DeepgramGroqCartesia Sonic-3Python AsyncIOPydanticsounddevice

FinDocMind

Nearing Completion

Enterprise-grade AI-Powered Document Intelligence Platform that transforms unstructured financial documents (PDFs, Excel, CSV) into structured Extraction Grids using a sophisticated three-stage AI agent pipeline built on OpenAI's GPT models.

Agent Architecture

1
Page Analyzer

Page-level field discovery with context-aware extraction

2
Doc Consolidator

Document-level field merging and deduplication

3
Multi-Doc Orchestrator

Cross-document field standardization and balancing

Highlights

  • Three-Stage AI Agent Pipeline with hierarchical orchestration
  • GPT-powered field discovery with semantic deduplication
  • Multi-format processing: PDF, Excel, CSV
  • Async concurrent processing pipeline with FastAPI
  • Type-safe Pydantic models, SQLAlchemy ORM, structured logging
OpenAI GPT-4LangChain AgentsFastAPISQLAlchemyPydanticPython AsyncIOPostgreSQL
View on GitHub

Things I've built

Big Data Stock Market Analysis

Processing and analysis of large-scale stock market data using distributed computing frameworks.

  • Real-time data streaming
  • Distributed computing at scale
  • Containerized deployment
Apache KafkaApache SparkHadoopDocker

Story Weaver

Interactive storytelling model using NLP to maintain context of evolving storylines in real-time.

  • Context-aware generation
  • RL for coherent continuations
  • Real-time interaction
NLPGPT-3Reinforcement LearningTensorFlowPyTorch

Parking CV System

Automated parking management using computer vision for real-time spot detection and plate reading.

  • Real-time spot detection
  • Vehicle recognition
  • License plate reading
OpenCVYOLOEasyOCRPython

Tech stack

Agentic AIPydanticAILangChainLangGraphCrewAIRAG PipelinesVector DatabasesKnowledge GraphsLLM OrchestrationPrompt EngineeringNLPMultimodal AIAgentic AIPydanticAILangChainLangGraphCrewAIRAG PipelinesVector DatabasesKnowledge GraphsLLM OrchestrationPrompt EngineeringNLPMultimodal AIAgentic AIPydanticAILangChainLangGraphCrewAIRAG PipelinesVector DatabasesKnowledge GraphsLLM OrchestrationPrompt EngineeringNLPMultimodal AI
TensorFlowPyTorchDeep LearningComputer VisionYOLOCNNsTransformersReinforcement LearningSnowflakeApache SparkKafkaHadoopTensorFlowPyTorchDeep LearningComputer VisionYOLOCNNsTransformersReinforcement LearningSnowflakeApache SparkKafkaHadoopTensorFlowPyTorchDeep LearningComputer VisionYOLOCNNsTransformersReinforcement LearningSnowflakeApache SparkKafkaHadoop
PythonJavaCJavaScriptTypeScriptSQLAzure Data FactoryAzure SynapseFastAPIPydanticSQLAlchemyPostgreSQLPythonJavaCJavaScriptTypeScriptSQLAzure Data FactoryAzure SynapseFastAPIPydanticSQLAlchemyPostgreSQLPythonJavaCJavaScriptTypeScriptSQLAzure Data FactoryAzure SynapseFastAPIPydanticSQLAlchemyPostgreSQL
DockerKubernetesLinuxGitHubShellReactNode.jsGitOpenCVEasyOCRNext.jsTailwind CSSDockerKubernetesLinuxGitHubShellReactNode.jsGitOpenCVEasyOCRNext.jsTailwind CSSDockerKubernetesLinuxGitHubShellReactNode.jsGitOpenCVEasyOCRNext.jsTailwind CSS

Certifications

Snowflake

Data Warehouse Badge

Snowflake

Data Engineering Badge

HackerRank

Java (Intermediate)

HackerRank

Python (Intermediate)

Infosys Springboard

Data Science

Infosys Springboard

Artificial Intelligence

AWS

Cloud Architect

AWS

Data Analytics

AWS

Data Engineering

Academic background

B.Tech in Computer Science (Data Science)

DJ Sanghvi College of Engineering, Mumbai

2024CGPA: 8.9

XII (HSC)

Pace Junior College, Mumbai

202083%

X (ICSE)

Shri Balaji International School, Mumbai

201896%

Let's build something
together.

Have an idea, a project, or just want to chat about AI and multi-agent systems? I'd love to hear from you.

© 2026 Tanay Agarwal

Designed & crafted in Mumbai