I build data‑driven products that turn complex datasets into clear, actionable insight—using modern analytics, machine learning, and careful experimentation.
Passionate about responsible AI, ML engineering, and credible causal inference.
A few representative works spanning predictive modeling, NLP, computer vision, and data engineering. Filter to explore.
Time‑Series Predictive Maintenance
Predictive Maintenance with LSTMs
Built an end‑to‑end pipeline to forecast equipment failures using sensor telemetry (PyTorch, MLflow, Airflow). Included feature store, drift monitoring, and CI/CD to deploy to an API.
Improved early‑warning horizon while reducing alert fatigue (replace with your metric).
Automated retraining & model registry with MLflow; tracked lineage and reproducibility.
Designed a privacy‑aware RAG system with dense retrieval and hybrid ranking; evaluated hallucinations with human‑in‑the‑loop labeling and prompt‑level guardrails.
Improved answer faithfulness and reduced unsupported claims (insert measured deltas).
Deployed on‑prem with vector DB; added anonymization and PHI redaction.
Implemented transfer learning for visual inspection; combined weak supervision with focal loss to handle class imbalance; deployed as a lightweight on‑edge model.
Cut manual review time and stabilized precision/recall across lighting conditions.