Insights
What we write and what we measure.
Engineering writing and applied research in one stream. ScaleJade works in the open, measures honestly, and publishes systems that hold up under real-world demands.
- PapersPreprint · 2026
TutorBench: Evaluating Large Language Models as Adaptive Tutors
A benchmark and evaluation suite for measuring how well LLMs teach — not just answer — across adaptivity, correctness, and pedagogy.
- PapersPreprint · 2026
LedgerProof: Verifiable Settlement for Permissioned Blockchain Networks
A settlement protocol that produces succinct, externally verifiable proofs of finality for permissioned ledgers used in regulated finance.
- PapersPreprint · 2026
DriftGuard: Detecting Silent Data Drift in Production ML Pipelines
A lightweight monitoring method that flags distribution shift in deployed models before it degrades downstream business metrics.