AI Engineer for Chemical Document Decoding: Unleashing Insights from PDFs
Objective
The objective of this project is to develop a solution that can read PDF documents and extract various types of information, including chemical structure drawings, spectra (NMR), boiling points, melting points, and synthesis information. Additionally, the solution should be able to combine core and variable parts of chemical structure images to generate multiple complete images. A prototype implementation is provided that can read PDFs with generic encoding, extract structured text, and extract complete images. The task is to link multiple image components into a complete image and link the images to structured text and unstructured text.
Requirements
a.PDF Document Parsing: The AI solution should be able to parse PDF documents with standard encoding or complex non-generic encoding, and extract text and complete images using the provided prototype implementation.
b. Image Component Linking: The solution should be capable of linking multiple image components to generate complete images. This involves combining the core part of a chemical structure image with the variable parts to generate multiple complete images.
c. Text-Image Linking: The AI model should be able to link the extracted images to the corresponding structured text, which is presented in paragraphs.
d. Unstructured Text Extraction: The solution should be able to identify and extract unstructured text that may appear between different images or in footnotes within the PDF documents.
Implementation Options
a. Fine-tune Prototype Implementation: The engineer can enhance the provided prototype implementation with capabilities to link image components, link images to structured text, and extract unstructured text.
b. Build Custom Implementation: Alternatively, the engineer can build their own implementation from scratch if they believe it will yield better results.
Technology Stack
• Programming Language: Python
• Machine Learning Frameworks: TensorFlow, PyTorch, or similar
• Image Processing Libraries: OpenCV, PIL, or similar
• Natural Language Processing Libraries: NLTK, SpaCy, or similar
• PDF Parsing Libraries: PyPDF2, pdfminer.six, or similar
Deliverables
a. Python code that implements the AI model for linking image components, linking images to structured text, and extracting unstructured text.
b. Detailed documentation explaining the architecture, algorithms, and usage of the solution.
c. Test cases for validating the solution.
This job is already closed and no longer accepting applicants, sorry.
