NLP for Business Intelligence
We built a Python NLP service to parse candidate resumes and score matching skills in 250ms.
cut the time spent screening applications by 60%.
service scores candidate relevance profiles in 250ms.
System Bottlenecks
Recruiters spent hours reading PDF resumes and text applications manually. The process was slow and created hiring delays.
Architectural Move
We built a Python service using spaCy to extract entities and match text vectors. The model runs on a FastAPI backend, matching resume keywords against job definitions using TF-IDF indexing.
Execution Path
Entity Parser: We wrote spaCy rules to parse and extract skill names from unstructured text.
Inference API: We set up FastAPI to receive document payloads and run model scoring.
User Portal: We built a Streamlit page where recruiters can drag in CSVs and PDF batches.
"We dragged in our entire applicant folder and had a ranked shortlist in minutes. The tool extracted technical skill keywords without misreading formatting columns."
— Saad Kamran, HR Director
Core Stack
- Python (FastAPI & spaCy model engine)
- Streamlit (User portal)
- Uv (Fast dependency installer)
Under-the-hood Challenge
Parsing layouts from PDF and Docx files without breaking text sequences. We solved this by writing a reader that matches text bounding boxes to reconstruct paragraph flows before passing them to the spaCy parser.
This project replaces manual text screening with automated parsing. By wrapping Python ML libraries inside a FastAPI server, we help teams classify resumes and customer feedback text streams in milliseconds.