Publications & Manuscripts

Manuscript drafts in self-supervised learning and real-time ML systems.

Entries below are complete manuscript drafts. None have been submitted to a journal, conference, or preprint server yet.

Self-Supervised Flood Precursor Detection on USGS Streamgage Records

Aditya Shrestha
Complete Draft · Not Yet Submitted
Investigates self-supervised temporal encoders for non-linear hydrological anomaly detection, surfacing precursor indicators of flood events directly from raw streamflow series.

Backpressure-Driven Elastic Feature Windows for Real-Time ML Pipelines

Aditya Shrestha
Complete Draft · Not Yet Submitted
Proposes a load-adaptive stream processing architecture that dynamically modulates sliding feature windows under backpressure to reduce processing delay under variable ingestion load.

Self-Supervised Anomaly Detection for Glacial Lake Outburst Flood Precursors via Satellite Time Series

Aditya Shrestha
Complete Draft · Not Yet Submitted
Presents a time-series masked autoencoder pipeline for remote monitoring of moraine dams, evaluated with a documented falsification test on South Lhonak lake data.