AI Intern

ManageArtworks (A flagship
product brought to you by Karomi, a leading Enterprise SaaS provider) enables
4000 leading Global and Indian brands. We offer everything to get artwork
projects going & manage every step of the packaging and artwork process.
Companies reach markets faster with our end -end packaging & artwork
management system while achieving 100% compliance.

Project Duration: 2
Months

Location: Chennai – On Site

• Develop a visual segmentation and grouping system to detect
  text blocks from packaging artwork PDFs.
  
• Extract text using PDF parsing, OCR, and vector
  path analysis.
  
• Build a classification module to categorize copy (e.g.,
  ingredients, nutrition, warnings, claims, dosage, legal statements).
  
• Use LLMs and multilingual embeddings for semantic
  understanding across multiple languages.
  
• Handle complex scenarios such as:
  
• Text in different layers or flattened PDFs
  
• Vectorized or dotted text
  
• Multi -language copy (EU languages, Indian languages, etc.)
  
• Mixed visual objects and symbols
  
• Develop logic to associate text with nearby icons (e.g., ℮ mark,
  Veg symbol, certification marks).
  
• Create a structured JSON output representing the extracted
  copy blocks.
  
• Build small visualization/debug tools to highlight extracted regions
  on artwork.
  
• Collaborate with domain experts in food, pharma, and cosmetics labelling.
  

Requirements

Technical Skills

• Strong Python fundamentals
  
• Experience with one or more of the following:
  
• pdfplumber, PyMuPDF (fitz), pdfminer
  
• OpenCV for image/geometry operations
  
• OCR tools (Tesseract, EasyOCR, TrOCR)
  
• Deep learning frameworks: PyTorch or
  TensorFlow
  
• NLP/LLM frameworks: HuggingFace
  Transformers, Sentence Transformers
  
• Multilingual embeddings (Indic, EU,
  global multilingual models)
  
• Document AI models like
  LayoutLM, DocTr, or LayoutLMv3 (bonus)
  

Analytical Skills

• Ability to reason about layouts, visual hierarchy, and grouping
  logic
  
• Interest in combining vision + language intelligence
  
• Attention to detail when dealing with regulatory copy
  

Nice -to -Have Skills

• Experience with YOLO, Detectron2, or Vision Transformers
  
• Understanding of packaging artwork concepts (die -lines, copy,
  compliance, symbols)
  
• Basic knowledge of food/pharma/cosmetics regulatory layouts
  
• Familiarity with multilingual text handling.
  

• Multimodal Document AI (vision + text)
  
• Advanced PDF engineering and vector processing
  
• Text extraction from complex packaging artwork
  
• Deep learning for layout detection
  
• Working with multilingual datasets
  
• Real -world packaging industry challenges
  
• Regulatory labelling requirements across categories
  

By the end of the internship, you will have built:

• An end -to -end copy extraction pipeline
  
• A visual grouping engine
  
• A multi -domain classification system
  
• Structured JSON outputs for downstream compliance and automation
  workflows
  
• A working demo/mini -tool to test extraction accuracy