Hardik Logo Image
Hardik
hamburger menu hamburger menu close

Hey, I'm Hardik — AI/ML Engineer

I build RAG pipelines, LLM agents, and multimodal chatbots. I care less about flashy demos and more about making generative-AI systems trustworthy — reranking, validation layers, and evaluation.

RESUME

About Me I'm an AI/ML Engineer based in Delhi, focused on retrieval-augmented generation, LLM agents, and multimodal systems. I build production-grade AI — from document ingestion and chunking to retrieval, reranking, and the validation layers that keep answers honest.

Get to know me!

I'm Hardik, an AI/ML Engineer based in Delhi and a recent B.Tech CSE graduate (MAIT Delhi). My work centers on retrieval-augmented generation: turning messy real-world documents into structured, queryable knowledge and serving it through answers you can actually trust.

On one flagship pipeline I processed 1.4M+ High Court documents with OCR and daily ETL, and I built MCP servers that let LLM agents use real tools. I care about the unglamorous parts — chunking, reranking, and validation — because that's what separates a demo from a system people rely on. Feel free to contact me here.

Contact

My Skills

Python
PyTorch
TensorFlow
Hugging Face
LangChain
LangGraph
RAG
LLMs
MCP
A2A
ChromaDB
Prompt Engineering
PostgreSQL
MSSQL
Azure
Docker
C++
MERN

Projects Here are some of my projects that I have worked on.

DocuFlow cover

DocuFlow

An enterprise RAG orchestrator that converts PDFs, DOCX, and images into structured Markdown before ingestion — so tables and layout survive into the chunks. Hierarchical chunking, BGE embeddings, ChromaDB, and a three-way validation layer (Gatekeeper, Auditor, Strategist) that keeps generated answers honest.

INSPECT
Judicial Pipeline cover

Judicial Pipeline

A legal-document pipeline that crawled High Court websites with Scrapy, OCR'd 1.4M+ documents, and loaded the cleaned metadata into a relational SQL database — with a daily Dockerized ETL on Azure, plus logging, monitoring, and automated alerts.

INSPECT
A2A Analytics ChatBot cover

A2A Analytics ChatBot

A conversational analytics platform for complex enterprise datasets. LangChain pipelines ground every answer in verified MSSQL data, and an adaptive Agent-to-Agent (A2A) framework routes between a fast path and a deep-research path.

INSPECT
MCP Server Suite cover

MCP Server Suite

A suite of Model Context Protocol servers that let LLM agents use real tools — Git version-control operations, Python linting & formatting, and a professional leave-management server. MCP is fast becoming the standard for wiring agents to tooling.

INSPECT
AI Voice Assistant cover

AI Voice Assistant

A real-time conversational voice agent for customer support, built on LiveKit WebSockets with AssemblyAI for speech-to-text, Cartesia for low-latency text-to-speech, and Silero VAD for accurate turn detection — so the conversation feels human, not robotic.

INSPECT
Vision Agent cover

Vision Agent

A multimodal AI agent that perceives its environment through vision — images and screenshots — and uses an LLM to reason and act. A hands-on step toward computer-use agents.

INSPECT
TeleBot cover

TeleBot

A conversational AI agent that lives inside Telegram — turning a familiar chat app into a surface for LLM-powered assistance.

INSPECT
Earlier work · Full-Stack
Flick Flair cover

Flick Flair

A movie review website where users can see reviews, ratings, and discover new movies — powered by the TMDB API.

INSPECT
GenG cover

GenG

A virtual assistant built with JavaScript and the Web Speech API — a conversational AI that understands voice commands and responds.

INSPECT
Auralify cover

Auralify

A music website where you can listen to and discover new music for free, anytime and anywhere.

INSPECT
Book Notes cover

Book Notes

A book-collection manager built with Node.js, Express.js, and PostgreSQL — view, search, and add books to your shelf.

INSPECT
LifeScript cover

LifeScript

A dynamic blog website built with EJS and Node.js, featuring real-time updates and an intuitive writing experience.

INSPECT

Contact Let's connect! Feel free to reach out to me for any project collaboration or just to say hi. I'm always open to new opportunities and looking forward to hearing from you.