AI Placement System
Led a four-person team as Product Lead to cut manual resume screening by 60% with an NLP pipeline.
The problem
A college placement cell was drowning in resumes every season. Coordinators manually read and shortlisted hundreds of candidates against role requirements — slow, inconsistent, and impossible to scale across recruiters. The bottleneck wasn’t a lack of candidates; it was human screening throughput.
How I approached it (as Product Lead)
This was my first time owning a product end-to-end with a team, and I leaned hard on process to keep four people aligned.
- PRD with RICE. I authored the product requirements doc and prioritized features by Reach, Impact, Confidence, and Effort — so we built the screening engine before the nice-to-have analytics.
- Three Agile sprints. I ran the team across three sprints with clear sprint goals, standups, and a shared definition of done.
- Quality as a gate. I defined 12 UAT test cases up front. Nothing shipped until defects were resolved — we hit 100% resolution before launch.
What we built
A resume-screening system that uses Google Gemini-Pro NLP to parse résumés, extract structured signals, and rank candidates against role criteria — wrapped in a Flask app backed by MySQL.
Outcome
- 60% reduction in manual resume-screening effort for coordinators.
- 30% efficiency gain in the overall placement workflow.
- 100% of defects resolved before launch, validated across 12 UAT cases.
What I learned
The hardest part wasn’t the model — it was alignment. Writing the PRD and prioritizing with RICE meant that when scope pressure hit mid-sprint, we had a shared, defensible answer for what to cut. Process is what let a small team ship confidently.