Diagnostic Signal Modeling & QC Automation

Role: Project PM / Data Scientist  ·  Stack: Python, R, LSTM, linear/basis-function modeling, PCA/t-SNE/DBSCAN, R Shiny

Two diagnostics projects where statistical rigor drove measurable safety and cost outcomes.

PCR signal baseline correction

• Redesigned a hard-coded legacy baseline algorithm into a data-driven mixed-basis-function model, cutting the false-negative rate 0.47% → 0.04% (91.49% improvement).
• Ranked #1 in residual-signal white-noise approximation against 5 competing algorithms; refactored Matlab → Python with real-time linear-regression optimization.

Equipment QC automation

• Replaced manual Excel QC with a two-stage LSTM + 10 graded performance metrics, cutting QC time ~93% (≈13× annual operating-cost reduction).
• Trained on 2,201 devices / 2,552 runs / 61,248 signals: pass-fail classification 94.5%, grade classification 82.7%; anomaly detection via PCA, t-SNE, DBSCAN, 3-sigma with an R Shiny real-time dashboard.
• Recognized with an R&D President’s Award and two first-inventor patents.

Approach

Both projects replaced a manual or hard-coded baseline with a measured, data-driven model.

flowchart TB
    subgraph PCR[PCR baseline correction]
      direction TB
      L[Hard-coded legacy algorithm] --> D[Data-driven<br/>mixed-basis-function model]
      D --> FN[False-negative<br/>0.47% to 0.04%]
    end
    subgraph QC[Equipment QC automation]
      direction TB
      M[Manual Excel QC] --> LSTM[Two-stage LSTM<br/>+ 10 graded metrics]
      LSTM --> T[QC time ~93% lower]
    end