Benjamin Piscoya

WORK HISTORY

EDUCATION

ABOUT BENJAMIN PISCOYA

Early‑career ML/AI practitioner with a strong foundation in time‑series forecasting, marketing‑mix modelling and end‑to‑end ML pipeline design. I’m completing an MSc in Data Science (UWE Bristol) and hold the MITx MicroMasters in Statistics & Data Science (IDSS). My recent work focuses on building leakage‑safe forecasting pipelines that combine tree‑based models (XGBoost) with LSTM residual correction, tuned via Bayesian optimisation (Hyperopt/TPE) and validated with rolling, time‑ordered cross‑validation and bootstrap uncertainty. I enjoy turning messy, multi‑source data into production‑ready features and models: GCP BigQuery/Dataflow/PubSub for ingestion, Python/pandas/scikit‑learn/PyTorch for modelling, and pragmatic MLOps (versioned artefacts, MLflow, Docker) for reliable deployment. I’m comfortable with feature engineering (lags, ad‑stock & saturation transforms, holiday flags), experiment design, and transparent evaluation (RMSE/MAE, skill vs baselines, planning‑horizon stability). What I bring: Applied ML on real commercial datasets (daily TV demand, price, supply, multi‑channel advertising) with measurable improvements vs naïve baselines. Marketing analytics → ML: features, causal thinking, and business‑ready dashboards to communicate model impact to planners and marketers. Data engineering mindset: reliable ETL, reproducible code, and documentation so models can be used, audited, and improved by teams. Now seeking roles such as ML Engineer, Data Scientist (ML/AI/Forecasting) or Junior MLOps Engineer in the UK/Ireland (open to hybrid/remote). I’m particularly interested in problems involving demand forecasting, pricing, ad‑effect measurement, recommender systems, and productionising ML. Tech: Python (NumPy, pandas, scikit‑learn, PyTorch, XGBoost), R, SQL, GCP (BigQuery, Dataflow, Pub/Sub), MLflow, Docker, Airflow, Plotly, Power BI, Tableau. Domain: Time‑series ML, experimental design, uncertainty quantification, ETL & data quality.