Python Developer PDF Table Extraction (Open-Source OCR & AI)

We are seeking a skilled **Python Developer** with expertise in extracting **unstructured tables from PDF documents** using **open-source models**. The ideal candidate should have hands-on experience with **OCR, deep learning, and NLP techniques** to accurately process and structure tabular data.  

### **Key Responsibilities:**  

- Develop and implement **Python-based solutions** to extract tables from **unstructured PDFs**.  

- Utilize **open-source libraries** like **pdfplumber, Tesseract OCR, Camelot, Tabula, PyMuPDF**, and deep learning-based models.  

- Handle **complex table structures, multi-page tables, and merged cells** effectively.  

- Preprocess PDFs, including **noise reduction, skew correction, and text enhancement**.  

- Use AI/ML models (e.g., **Detectron2, LayoutLM, Donut OCR, or Graph Neural Networks**) for intelligent table extraction.  

- Optimize the accuracy and reliability of extracted data through **post-processing techniques**.  

- Ensure **scalability, performance, and error handling** for large document processing.  

- Work with **structured storage solutions** like **Pandas, SQL, or JSON** for extracted data.  

- Collaborate with teams to **integrate the solution into an existing pipeline or API**.  

### **Required Skills:**  

✅ **Strong Python skills** (NumPy, Pandas, OpenCV, TensorFlow/PyTorch).  

✅ **Experience with OCR tools** (Tesseract, EasyOCR, PaddleOCR).  

✅ **PDF processing libraries** (pdfplumber, PyMuPDF, Camelot, Tabula).  

✅ **Deep Learning models** for document understanding (Detectron2, LayoutLM, Donut OCR).  

✅ **Preprocessing techniques** (denoising, deskewing, contour detection).  

✅ **Experience with NLP and Computer Vision for text segmentation**.  

✅ Knowledge of **data extraction, transformation, and validation techniques**.  

✅ Familiarity with **Docker, API integration, and cloud storage solutions**.  

### **Preferred Skills (Bonus):**  

🔹 Experience in **Graph Neural Networks (GNN) for table structure detection**.  

🔹 Working knowledge of **Hugging Face transformers for document AI**.  

🔹 Familiarity with **LLMs for intelligent document parsing (LlamaIndex, LangChain)**.  

### **Project Goal:**  

Develop an **end-to-end open-source solution** that accurately extracts and structures tables from **scanned and text-based PDFs** without using paid services like **AWS Textract, Google Vision, or Azure Form Recognizer**.