Research

Selected projects and applied research experience

This section is designed as a living overview of what I have explored, built, and validated in practice.

Object Detection for Real-World Creative Safety

2023

Explored DETR-family detectors, ViT-based feature pyramids, and production evaluation for detecting policy-violating elements in ad imagery.

  • Compared YOLO, Co-DETR, and Co-DINO variants against internal benchmarks.
  • Adapted pre-trained ViT backbones to dense detection with multi-scale feature construction.
  • Observed practical gains in online A/B evaluation despite modest offline metric deltas.

From Software Engineering to Applied ML

2020-2022

Moved from software engineering into algorithm engineering, with a focus on production-ready ML systems and disciplined experimentation.

  • Built engineering habits around reproducibility, iteration speed, and debugging complex pipelines.
  • Turned research papers into testable implementation plans instead of purely theoretical reading.