Building production-grade ML systems that ship — from multimodal transformers and RAG pipelines to real-time AWS inference APIs. M.S. Data Science & AI · UCM · 4.0 GPA.
I don't just build models — I ship ML systems. My work spans the full stack: statistical research through model training, containerized deployment, REST APIs, and CI/CD automation on AWS.
Currently an M.S. student in Data Science & AI at the University of Central Missouri with a perfect 4.0 GPA, I spend my time building production RAG pipelines, multimodal deep learning systems, and real-time fraud detection APIs you can actually hit with a curl command.
I'm looking for full-time roles in Data Science, ML Engineering, or Applied AI in the United States — available June 2026.
Production-structured RAG pipeline answering questions over 120 arXiv ML papers. Rigorously benchmarks BGE, MPNet, MiniLM vs BM25 across chunk sizes and generation quality. BGE achieves perfect MRR 1.000 — 7.7× Answer Relevance over BM25 (0.910 vs 0.118).
End-to-end production ML system. Automated GitHub Actions CI/CD trains, evaluates, and promotes models to MLflow Model Registry (PostgreSQL RDS). Dockerized FastAPI inference API behind Nginx on AWS EC2. XGBoost & Random Forest hit ROC-AUC 1.000 with SMOTE-balanced pipeline.
Fuses text (BERT), audio (Wav2Vec2), and video (ViT) using a custom 4-layer Transformer Encoder (8 attention heads) for 7-class emotion classification on MELD. Mixed-precision AMP training, staged fine-tuning, cached tensor preprocessing, and class-weighted loss for minority emotion handling.
Systematic comparison of TF-IDF + Logistic Regression vs fine-tuned DistilBERT on 62K+ articles. Dynamic F1-threshold optimization, NLTK preprocessing, feature explainability from TF-IDF weights, and real-time NewsAPI integration for live article classification via Streamlit.
End-to-end epidemiological forecasting: live JHU data ingestion → preprocessing → feature engineering → multi-model benchmarking (SARIMAX, Prophet, XGBoost) → interactive Plotly Dash dashboard with 30-day forecasts and 95% CI bands. Seasonal SARIMAX(1,1,1)(1,1,1,7) wins with Holdout MAE 138.14 ≈ CV MAE 137.86 — near-zero overfitting confirmed across 5-fold rolling cross-validation.
Actively seeking full-time Data Science, ML Engineering, and Applied AI roles in the United States. Available to start June 2026. Let's build something that matters.