Deven Varu

Applied LLM automation / Pistachio Internship

Catalog Intelligence Automation

AI pipeline that turns 20-200 page furniture catalogs into structured product data and linked product imagery.

Pistachio internship / Software engineer intern and team lead / Shipped internally

OpenAI / OCR / YOLO / JSON Extraction / Python / Team Lead

Catalog Intelligence Automation Project screenshot coming soon

Highlights

  • Led a three-person team replacing manual catalog data entry with an AI extraction workflow.
  • Tested Gemini, OpenAI, Claude, and OCR tools before standardizing on OpenAI for reliable JSON output.
  • Added YOLO image detection to extract and connect catalog visuals with product records, reducing work from hours to minutes.

Context

During the Pistachio internship, the project focused on replacing manual furniture catalog data entry for 20-200 page catalogs.

What I built

  • A three-person team workflow for AI-assisted catalog extraction.
  • LLM and OCR evaluation across Gemini, OpenAI, Claude, and OCR tools.
  • Structured JSON extraction standardized on OpenAI for reliable output.
  • YOLO image detection to extract and connect catalog visuals with product records.

Result

The workflow reduced catalog processing work from hours to minutes for the supported internal use case.

Back to projects

Related projects

  • Road Asset Detection with YOLO - Computer vision pipeline for Boone County that detects and maps road infrastructure from stitched 360-degree roadway imagery.
  • Here2Stay Interest Form - Bilingual web portal for EBPREC that digitized community intake, sent resource packets, and answered housing-resource questions with a RAG chatbot.
  • Cisco HackAIThon Project - Natural-language analytics chatbot that translates user questions into SQL and surfaces device usage and memory insights in a dashboard.