AI for Job-Description Classification

A NASDAQ-listed staffing and professional-services firm engaged Taller to build the AI models and data pipeline behind its job-description classification capability.

The client’s competitive-intelligence function wanted to know which companies were posting jobs on staffing-agency sites under generic descriptors — an intelligence exercise that would sharpen the firm’s competitive positioning across technology and financial-services placements. The challenge was building the data pipeline and classification models at production scale, not as a one-off research experiment. The client engaged Taller to deliver both.

Taller built and trained AI models against scraped job-description content, sorting roles by sector and extracting the skills each posting named. Scrapy and Selenium handled the data harvesting (Scrapy crawls sites at scale, while Selenium drives a real browser to reach content that only loads through user interaction). Pandas and Jupyter Notebook supported the data-science work, giving the team a workspace to explore, clean, and reshape the data as they refined the models. TensorFlow powered the model training itself. What separated a production-grade pipeline from a research experiment was the classification taxonomy: sector-by-sector rules that matched how the client’s commercial team actually segmented the market, with the model’s confidence scores wired into the recruiter-facing tool so low-confidence classifications got human review before influencing outreach.