Story Tag
I am currently pursuing a PhD in Computer Science at the University of Louisiana at Lafayette, where my research focuses on internet cybercrime, Web3 and internet security, large-scale web measurement, phishing, abuse, and scam ecosystems. During my prior MS in Computer Science at the University of New Orleans, I specialized in machine learning and deep learning for computer vision. I developed robust pipelines for real-time segmentation, object detection, and annotation automation, significantly improving efficiency and accuracy in levee fault detection. My work also included building scalable full-stack pipelines for real-time inference, enhancing dataset diversity with diffusion models, and applying model quantization techniques to improve efficiency. I am a lifelong learner and an enthusiastic collaborator who enjoys turning research ideas into practical, automation-driven security systems.
Tools and technologies I use to build robust, secure AI systems from research to deployment.
A short snapshot of my academic background and industry engineering work.
Strong academic foundation in computer science, with ongoing doctoral research and prior graduate specialization in AI/ML.
Hands-on full-stack engineering experience building reliable, scalable systems for production use.
Source-backed global view of campaign clusters, abuse infrastructure, and victim-origin patterns with live rotating context highlights.
Live snapshot: loading...
active campaign clusters tracked this quarter
sinkholed and monitored suspicious domains
infrastructure overlap signals under active review
Regional distribution for the selected view, driven by the loaded threat dataset and public-source references.
Validated references for the active view:
A compact view of production-minded ML workflows with hands-on demos.
Scope risk and define measurable success criteria.
Artifact: Success Metric SheetAssemble and validate representative training data.
Artifact: Dataset Snapshot and Label AuditRun controlled experiments with tracked configurations.
Artifact: Experiment Run LogStress-test for precision, recall, and failure modes.
Artifact: Confusion Matrix and Error BucketsPackage model service with latency-aware API paths.
Artifact: API Latency and Throughput ReportTrack drift and trigger safe retraining workflows.
Artifact: Drift Dashboard and Alert RulesDefine task scope, operating constraints, and measurable outcome targets before training begins.
Primary Artifact: Success Metric SheetBuilt for the same style of end-to-end CV workflows used in segmentation and detection projects from MS research and production engineering practice.
Use one scene across segmentation, explainability, and synthetic augmentation. Switch sub-views below without leaving this panel.
Semantic mode: one foreground class for geometric objects over background.
Ready. Choose a mode and click Run Segmentation.
Inspect prediction-vs-ground-truth drift, saliency, and boundary stress on the same geometric segmentation scene.
Ready. Rendering explainability overlays on geometric scene.
Preview hard-case augmentations directly on the same segmentation scene and estimate generalization gain.
Choose a transform to preview how it affects the segmentation scene patch.
Baseline mIoU: 0.68 | Boosted mIoU: 0.77 | Generalization Gain: +0.09
Ready. Live augmentation preview is active.
A practical decision guide: choose an algorithm by problem shape, data volume, and explainability needs, with live examples teams can quickly map to real product scenarios.
Live scenario mode: algorithm behavior, split quality, and diagnostics update continuously.
Scenario: binary conversion risk scoring with constrained latency and explainability needs.
Switch algorithms to see when each one should be used in a delivery workflow, not just how it scores.
Visualize how each sorting strategy transforms an array over time and compare implementation trade-offs beyond just Big-O notation.
Scenario: choose a stable sorter for transaction feeds where duplicate keys must preserve upstream ordering.
Ready: choose an algorithm, generate data, and run to inspect each strategy step-by-step.
A practical delivery loop blending MS research habits with production engineering discipline.
Understand problem boundaries, users, constraints, and risk. Translate requirements into measurable success criteria before implementation.
Output: problem framing doc, measurable KPIsBuilt malware risk prediction using ensemble learning and compared model families for practical detection workflows.
Applied imbalance-aware training strategies to improve fraud identification reliability in realistic transaction datasets.
Security research code and experiments around domain abuse tracking and suspicious infrastructure behavior.
Crypto ecosystem analysis project combining PySpark pipelines with predictive modeling.
End-to-end deep learning deployment for levee deficiency analysis with practical controls for field usage.
Multi-class traffic sign recognition pipeline designed for autonomous driving-oriented scenarios.
Clinical image classification project with class imbalance mitigation and explainable performance reporting.
Emotion classification benchmark with precision/recall focused evaluation by class.
AI-powered stock screener and signal engine for Nepal's NEPSE exchange with live dashboards and ML buy/sell signals.
Exploratory analytics platform to extract and communicate insights from large spatiotemporal ocean datasets.
Business intelligence dashboard for supply-chain visibility and faster strategy decisions.
Text analytics pipeline for clustering and semantic proximity insights across corpora.
Neural machine translation pipeline with tokenization, sequence modeling, and bilingual evaluation workflows.
Hands-on NLP repository with preprocessing, feature engineering, and model experimentation tasks.
Language generation mini-project focused on transition-based stochastic text synthesis.
Word search and lexical matching project with performance-aware board traversal logic.
Autonomous navigation study combining planning and environmental mapping for robust UGV movement.
High-performance solver leveraging backtracking and multi-threaded parallelism for faster complex puzzle resolution.
Practical edge AI deployment repository focused on latency, model size, and device execution constraints.
Advanced algorithmic implementations and comparative ML experimentation workflows.
New Orleans, Louisiana, United States