Personal Project: Job Discovery & Auto-Apply Platform (Hand-labeled evaluation dataset for model benchmarking)
Don't disclose
Developed and evaluated an AI job discovery system using a custom classification model and retrieval pipeline. Hand-labeled a 100-example evaluation dataset and used implicit user signals as additional supervision for benchmarking. Produced benchmark scripts and evaluation dashboards to support model evaluation and threshold sweeps. • Hand-labeled 100 evaluation examples. • Captured implicit user signals (favorites and click-throughs) as labels. • Built benchmark scripts and Grafana dashboards for evaluation. • Ran model evaluation and threshold sweeps.
2025 - Present