📄 Paper · 💻 Code · 🌐 Portfolio · ✉️ Email · 💼 LinkedIn
- SADE-NetworkAgent — Symptom-Aware Diagnostic Escalation for LLM-based network troubleshooting: a phase-gated diagnostic workflow + 15-skill library that lifts root-cause F1 by +37 percentage points over a ReAct + GPT-5 baseline on NIKA's 523-incident test set. (IEEE LCN 2026 · arXiv:2605.04530)
- NIKA — contributor to the network arena for benchmarking AI agents on network troubleshooting; the SADE agent is integrated upstream.
- nem-power-data-pipeline — streaming pipeline for Australia's electricity market: REST acquisition → DuckDB star schema → MQTT replay → live Streamlit map dashboard (716k observations, 356 facilities). Data engineering.
- au-energy-data-pipeline — ETL that scrapes public Australian energy & emissions open data (NGER, Clean Energy Regulator, ABS) into a spatial DuckDB warehouse. Data engineering.
- cnn-image-classifier — CIFAR-10 benchmark of Random Forest vs MLP vs a VGG-style Keras CNN on identical splits — the CNN reaches 84.2% test accuracy vs ~50% for the flat-input baselines. Deep learning.
- used-car-price-prediction — Linear Regression, Decision Tree & Random Forest (R² 0.91) with a full regression-diagnostics suite. Machine learning.
- airline-review-satisfaction — passenger-satisfaction classification on ~22k Skytrax reviews: five models benchmarked with/without SMOTE, leak-free pipelines (RF+SMOTE macro F1 0.693). Machine learning.
- car-dealership-db — a normalized PostgreSQL database — schema, ER model, sample queries, and a Flask demo app. Databases.
- neetcode-submissions — data structures & algorithms practice, in Python.
Python · LLM agents · Claude Code · PyTorch / Keras · scikit-learn · data engineering (DuckDB, ETL) · PostgreSQL · network diagnostics · Linux · Git
📫 Open to opportunities — jeffqaz741@gmail.com