Statistical Programming Leader & Data Scientist — Clinical Trials · Machine Learning · AWS AI/ML
I lead Biostatistical Programming at Amgen and I'm an active practicing Data Scientist. 12+ years in clinical research — SAS, CDISC (SDTM/ADaM), Pinnacle21, FDA / EMA / PMDA / CFDA submissions — plus a growing portfolio of Python ML work (pandas, scikit-learn, TensorFlow/Keras, XGBoost, MLflow, GitHub Actions CI). The same rigor that ships a regulatory submission is what ships a trustworthy model.
- 🎓 AWS Certified AI Practitioner, currently completing AWS Certified Machine Learning Engineer – Associate.
- 🧪 SAS Certified Base Programmer (SAS 9); co-author on 18 peer-reviewed publications; primary SAS programmer on five Phase 3 oncology trials at BMS.
- 🛠 Daily stack: SAS · Python · pandas · scikit-learn · TensorFlow/Keras · MLflow · Amazon SageMaker · Streamlit · GitHub Actions.
- 📜 CDISC SDTM/ADaM · Pinnacle21 · SAP/TFLs · GxP · ICH-GCP.
- 📍 Greater Tampa Bay Area, FL · open to remote.
- 🔗 LinkedIn · ✉️ bnelsontorres@gmail.com
End-to-end reproducible ML research pipeline on 1-minute OHLCV data for
44 Nasdaq-100 tickers (≈ 21.2 M rows, Nov 2020 – Dec 2025). Leakage-safe
feature engineering (62 features across 8 families), PCA + classical
baselines, Conv1D CNN, SimpleRNN / LSTM / GRU, MLflow experiment
tracking, TensorBoard, pytest + GitHub Actions CI, model cards,
and a Streamlit dashboard.
Python TensorFlow scikit-learn XGBoost LightGBM MLflow Streamlit
Hands-on Amazon SageMaker labs for the MLE-Associate exam: Spark
feature engineering on EMR, built-in + script-mode XGBoost training with
HyperparameterTuner, SageMaker Pipelines + Model Registry, and
Model Monitor / Data Capture for drift detection. Includes
modernization of legacy SageMaker SDK notebooks to sagemaker>=2.245.
AWS SageMaker boto3 EMR/Spark XGBoost MLOps
- Regulated-data discipline: CDISC (SDTM/ADaM), Pinnacle21, GxP, ICH-GCP — the same rigor MLOps requires.
- Reproducibility & QC at the level of FDA / EMA / PMDA / CFDA submissions, applied directly to ML pipelines (leakage-safe features, embargoed splits, pytest coverage, CI).
- Cross-functional leadership of geographically distributed (UK, China, India) programming teams.
- Comfort communicating model and study results to non-technical stakeholders — I've translated statistics for clinicians and regulators for over a decade.
This README lives at https://github.com/bentor79/bentor79 and renders on my profile.