← Atlas/Sheet 01/AI/ML · Open Source

FairLint DL

Deep-learning fairness debugger shipped as a VS Code extension

Role

Sole engineer

Timeline

Jan 2026 — May 2026

Year

2026

Category

AI/ML · Open Source

Interactive demo— simulated, runs in your browser

Survey notes

A deep-learning-based fairness debugger that ships as a VS Code extension. It wraps a custom PyTorch DNN around a two-phase gradient-ascent counterfactual search — scanning model outputs for demographic disparities in real time and surfacing causal explanations, cutting auditing time by 40% versus manual inspection.

Notable terrain

  • 01Real-time fairness scanning integrated into the VS Code sidebar
  • 02Two-phase gradient-ascent counterfactual search to isolate causal bias sources
  • 03Custom PyTorch DNN for demographic disparity detection across 12 fairness metrics
  • 04Interactive highlight layer — flagged rows glow directly in the editor
  • 0540% reduction in auditing time vs. standalone script-based workflows

Next sheet

relfair