Writing is how I pressure-test ideas, document tradeoffs, and translate research into production
decisions. These posts are structured like internal design reviews and research memos.
Project Deep Dive
Designing equity prediction models that survive data drift
A walkthrough of how I set up evaluation, leakage checks, and monitoring for alternative data
signals in production research pipelines.
- Problem framing and signal selection
- Validation strategy for non-stationary data
- Engineering guardrails to prevent silent failure
Drafting
Research → Practice
Calibration is not optional in financial classification
Translating neural network calibration research into decision-making thresholds for risk-aware
models.
- Why accuracy hides confidence issues
- Temperature scaling for deployment
- Aligning prediction confidence with action
Outline in progress
Technical Opinion
Most ML pipelines fail silently — here is how I audit them
A practical checklist for monitoring data contracts, model drift, and feature integrity in
production ML systems.
- Pipeline health as a first-class metric
- Detecting leakage and training skew
- Designing actionable alerts
Drafting
Want deeper dives? Reach out and I can share full drafts.
