- Parsing resumes was the hardest part, since they come in multiple formats (PDF, Word, scanned images) and contain unstructured data. Ensuring accurate extraction of skills, experiences, and certifications required fine-tuning NLP pipelines.
- Building reliable candidate-job fit scores was challenging, as it required balancing keyword matching with semantic understanding.