Featured Projects

AI systems built to solve real-world business challenges.

AI AgentsData PipelineWeb Scraping

AI-Powered School Intelligence Platform

EdTech Startup

Challenge

School information was scattered across thousands of websites—no structured, searchable dataset covering schools across India.

Approach

Built AI agents to discover, extract, validate, and structure data for over 13 million schools, creating a scalable intelligence platform for AI-powered search.

Outcomes

Centralized, AI-ready dataset covering 13M+ schools
Foundation for intelligent school discovery
Enabled future parent-focused search experiences

Tech Stack

PythonLangChainFastAPIPostgreSQLStreamlit

RAGKnowledge GraphMultimodal Search

Enterprise Knowledge Management System

Construction & Engineering Firm

Challenge

Critical business knowledge was scattered across PDFs, images, DWG drawings, videos, audio recordings, and spreadsheets—making information nearly impossible to locate.

Approach

Built a multimodal knowledge platform using RAG, Neo4j Knowledge Graphs, OCR, and Whisper to retrieve accurate answers with source citations across every document type.

Outcomes

Teams search years of engineering knowledge through natural language
Trustworthy, citation-backed responses
Eliminated hours of manual document searching

Tech Stack

PythonNeo4jOCRWhisperRAGFastAPI

Fine-TuningDomain Specialization

Fine-Tuned Llama 3.2 for Supply Chain Finance

Supply Chain Finance Company

Challenge

General-purpose language models lacked the domain expertise needed for finance-specific business queries, document summarization, and SQL generation.

Approach

Fine-tuned Llama 3.2 3B using LoRA on a custom Supply Chain Finance instruction dataset to improve domain understanding and response quality.

Outcomes

Lightweight, domain-specific AI assistant
Finance-focused Q&A, summarization, and SQL generation
Improved accuracy on finance-specific queries

Tech Stack

Llama 3.2 3BPEFTLoRAPython

Let's build something together.

Have a problem worth solving? I'd love to hear what you're working on.

Read the Blog