AI Multi Format Insight Engine
Budget: $750 – $1,500 AUD
I need a single AI-powered application that can take a batch of 500+ files—PDFs, Word docs, Excel sheets and mixed image formats—ingest them into its own searchable memory, and then let me pull meaningful intelligence from that content on demand.
Here is what I expect the system to do once the files are uploaded:
• For every text-based document it should automatically extract keywords and phrases, generate concise summaries with key insights, and run sentiment analysis.
• For images it must perform OCR to capture any embedded text and detect prominent objects that could add context to the overall analysis.
• All extracted information has to be stored internally so that I can query it later, ask follow-up questions, or run comparisons across the full data set without reprocessing the files.
A lightweight dashboard or command-line interface that lets me upload the files, track processing progress, and pose questions against the internal knowledge base would be ideal. You are free to recommend the tech stack—Python, JavaScript, vector databases, transformers, vision models, or any combination—as long as the final solution runs locally or on a private server I control and scales smoothly beyond the initial 500-file target.
When you deliver, I will review it against these acceptance criteria:
1. Successful ingestion and persistent storage of at least 500 mixed-format files.
2. Accurate keyword extraction, summary generation, and sentiment scoring for text documents.
3. Reliable OCR and object detection on images.
4. Interactive querying that returns responses in under five seconds for common requests.
5. Clear setup instructions so I can redeploy or fine-tune the model in the future.
If you have questions about file samples, preferred libraries, or hosting environment, let me know and I can provide additional details.
Here is what I expect the system to do once the files are uploaded:
• For every text-based document it should automatically extract keywords and phrases, generate concise summaries with key insights, and run sentiment analysis.
• For images it must perform OCR to capture any embedded text and detect prominent objects that could add context to the overall analysis.
• All extracted information has to be stored internally so that I can query it later, ask follow-up questions, or run comparisons across the full data set without reprocessing the files.
A lightweight dashboard or command-line interface that lets me upload the files, track processing progress, and pose questions against the internal knowledge base would be ideal. You are free to recommend the tech stack—Python, JavaScript, vector databases, transformers, vision models, or any combination—as long as the final solution runs locally or on a private server I control and scales smoothly beyond the initial 500-file target.
When you deliver, I will review it against these acceptance criteria:
1. Successful ingestion and persistent storage of at least 500 mixed-format files.
2. Accurate keyword extraction, summary generation, and sentiment scoring for text documents.
3. Reliable OCR and object detection on images.
4. Interactive querying that returns responses in under five seconds for common requests.
5. Clear setup instructions so I can redeploy or fine-tune the model in the future.
If you have questions about file samples, preferred libraries, or hosting environment, let me know and I can provide additional details.