Portfolio

Generates a synthetic 8-week chiller plant dataset with three embedded operational faults for diagnostic training.

Data quality checks, descriptive statistics, time-series visualization, and rolling SMA trend analysis of chiller plant operations.

Simple and multivariable linear regression with residual diagnostics and short-window prediction generalization.

Differential pressure analysis and Gaussian Mixture Model clustering to identify two distinct operational regimes.

Deep dive into low-delta-T syndrome using quantile-based thresholds and efficiency classification.

Heat rejection calculations, tower approach trending, fan power analysis, and free-cooling feasibility assessment.

Subsystem dominance analysis, energy waste quantification, and rolling SMA dashboards for final diagnostic conclusions.

Builds a compressor performance model for a grocery cold storage reciprocating rack system on R-448A and runs a temperature bin analysis using TMY weather data to estimate annual compressor energy savings from floating head pressure control under two retrofit tiers.

Demonstrates regression-based measurement and verification (M&V) for floating head pressure savings using simulated pre/post compressor power data, and performs sensitivity analysis on key assumptions including electricity rate, minimum condensing floor, and condenser approach temperature.

Fetches 15 years of IESO hourly demand data and Open-Meteo weather for Toronto, merges datasets, and performs comprehensive EDA on peak demand patterns.

Transforms raw demand and weather data into ML-ready features including humidex, cooling degree hours, demand momentum, and peak context variables.

Trains and compares regression and classification models for daily max demand prediction, with SHAP-based feature importance analysis.

Walk-forward backtesting across multiple base periods with confusion matrices, alert calendars, and financial valuation for a hypothetical 10 MW Class A customer.

Demonstrates a live prediction workflow with current IESO data and weather forecasts, and documents the REST API design and deployment architecture for operational use.

Ingests aggregated residential gas consumption by postal code, regional HDD data, and MPAC property tax roll summaries. Performs EDA on gas-temperature relationships across 150 synthetic postal codes.

Runs OLS regression per postal code, applies quality filters, normalizes slopes by MPAC building stock characteristics, and produces a thermal intensity metric for neighbourhood ranking.

Ranks postal codes by normalized thermal intensity within structure type groups, identifies priority neighbourhoods for retrofit incentive targeting, and tests sensitivity to normalization assumptions.

Generates realistic synthetic SCADA data for multiple WWTP service areas with distinct I&I severity profiles — including rainfall-driven RDII, snowmelt-driven infiltration, conductivity dilution, and pump station sub-catchment flow.

Establishes dry weather flow baselines using STL decomposition, computes RDII for each wet-weather and snowmelt event, and validates results against MECP per-capita flow guidelines.

Computes pre-whitened cross-correlation between rainfall/snowmelt and sewer flow for each plant, identifies dominant lag signatures, and classifies plants as inflow-dominated, infiltration-dominated, or mixed.

Trains XGBoost and LSTM models for RDII prediction, performs autoencoder-based anomaly detection for I&I period classification, and validates against conductivity-based dilution analysis.

Quantifies the annual energy cost of I&I at each plant, evaluates CIPP and private lateral remediation scenarios using cost-benefit analysis, and produces a composite severity ranking for capital planning prioritization.