Projects
From environmental sensing to low-latency infrastructure: systems, algorithms, and applied machine learning.
Research Projects
flood-sentinel
Complete Draft · UnsubmittedSelf-supervised temporal encoder trained on historical USGS streamgage data to detect precursor dynamics in environmental streamflows.
Manuscript summaryBPFeat
Complete Draft · UnsubmittedBackpressure-driven elastic feature windows for real-time ML streaming pipelines, designed to bound per-batch latency under variable ingestion load and evaluated against baseline streaming pipelines under a fixed test protocol.
Manuscript summarysentinel-gl
Complete Draft · UnsubmittedSelf-supervised masked autoencoder for detecting glacial lake outburst flood precursors from multispectral satellite time series, featuring a pre-registered falsification protocol and documented negative test bounds.
Manuscript summaryTopological Collusion Detection
Collaborative ResearchApplying topological deep learning and persistent homology computation to uncover collusion rings in financial transaction graphs across the Elliptic++ and IBM AMLworld benchmarks.
Engineering Systems
BankSentinel
1st Place · Global IME AI/ML Hackathon 2026A five-agent intrusion detection and behavioral transaction analysis system built with Team Eigen Threat and evaluated live before financial sector leadership.
KLStream
C++17 · Systems ProjectStream-processing runtime utilizing adaptive windowing for anomaly detection, featuring PID controller variants, LOBSTER ticker evaluation, and nanosecond timestamp accuracy.
ClickStream
Continuous BiometricsContinuous behavioral authentication system leveraging keystroke and mouse dynamics evaluated over a consented participant dataset.
CoreMesh
Distributed SystemsDistributed communication framework for resilient peer-to-peer data coordination under intermittent network conditions.
Load-Adaptive IIR Filtering
Digital Signal ProcessingAdaptive IIR filter design applying discrete Fourier transform and spectral decomposition algorithms to dynamic filtering challenges.