Specialization in data science with 8+ years of
experience in fast-paced, high-stress construction
environments as part of interdisciplinary project
management teams. Pursuing a professional career in data
science to find efficient, user-friendly solutions to
management systems in industries that are slow to adopt
new technologies.
As an unrepentant hobbyist, data science allows me to
develop tools that improve upon my existing projects as
well as open opportunities for new ones.
About Me
Education
Web Development
Udemy, 2023
Projects : Portfolio Website
Data Science
Lighthouse Labs, 2022
Projects : Mineral Exploration & Flight Delay Prediction
Civil Engineering
Western University, 2011
Experience
Engineering Research Lead Volunteer
Healthcare Systems, December 2023 to Present
- Continuing ongoing research from previous term.
- Lead on developing new model architecture.
Projects : Mineral Exploration
Engineering Research Assistant
Healthcare Systems, September 2023 to December 2023
- Assisted in developing machine learning models trained on magnetic map data intended to predict mineral concentrations for a given area.
- Developed tools to analyze model performance that directed future experiments towards incremental improvements.
-
Experiments included transfer learning models
trained on
imagenet
andbigearthnet
datasets. - Previous attempts at modeling a neural net trained on the magnetic map data resulted in \(R^2\) scores below 0; by term end, model performance saw positive scores.
- Owns and maintains submodule repository and documentation of development code.
Projects : Mineral Exploration
Freelance Data Analyst
Self-employed, October 2022 to Present
- Working with clients to retrieve and manipulate data accumulated over the last decade.
- Performed discoveries in the context of their business goals to uncover actionable insights.
- Directly communicated with COO to develop hypotheses and outline the investigation.
Project Coordinator
Four Seasons Site Development, March 2014 to July 2022
- Implemented and enforced production-specific safety programs and maintained clean safety records year-on-year in concert with clients, consultants, subcontractors, and company health & safety management personnel; challenging safety conditions include 30-m depth excavations and micro-tunneling underneath a provincial highway.
- Developed skills to meet challenging deadlines and negotiate unforeseen complications in the pre-construction and exploratory phases of most projects; client needs were met with minimal loss to time and budget.
- Created technical documentation for clients to satisfy contract requirements when closing projects.
Cadet Engineer
Stateland, April 2013 to February 2014
- Assessed BoM for semi-detached housing to spot deviations from historical data and/or implement changes according to design revisions to inform future design decisions relating to subdivision developments and other low-cost residential projects.
- Provided material take-offs to develop construction schedules and plans for material procurement.
- Consulted with vendors to closely monitor their efforts (and re-establish deadlines as necessary) and to conduct trade buy-ins for masonry works, steel framing, and for various architectural elements such as windows and doors.
- Facilitated meetings between senior engineers and vendors to negotiate costs, determine product specifications, and maintain collaborative relationships between all parties.
- Created and modified existing templates to automate routine assignments, which doubled work efficiency as well as enhanced accuracy and consistency by removing sources of human error.
Projects
Flight Delay Prediction
Data Science, Lighthouse Labs Capstone
Dataset Features
- Data & time of arrival
- Data & time of departure
- Airline name & airport of arrival and departures
- Plane configuration, capacity, and passenger count
- Flight distance & direction
Performed exploratory data analysis to inform data
processing and augmentation based on distribution,
biases, and counts using
scipy
and statsmodels
. Used
pandas
, numpy
, and boosting
techniques to process and normalize data, and then
applied PCA for feature reduction. Interfaced with
weather and traffic API to engineer new features.
Developed XGBoost model to predict flight delays and
tested against 2020 data.
Resulting \(R^2\) scores were poor. Further progress
could be made by incorporating data from services
affected by COVID.
Mineral Exploration
Healthcare Systems R&A
Dataset Features
- Magnetic map data
- Distance to volcanic formations
- Distance to faults
- Distance to outcroppings
- Adjacent mineral concentrations from rock samples
- XY-coordinate information and geodata for POI
Onboard onto existing project. Tasked with creating CNN
model using magnetic map data as feature dimensions.
Developed techniques to circumvent issues with the
dataset such as sparseness and low observation counts.
Preprocessing involved filtering for outliers (using
IQR) and null values. Normalized data using logarithmic
transformation (directly through the target variable or
indirectly by using a logarithmic loss metric). Images
were augmented using flips and rotations via the
OpenCV
library to boost observation counts
and to also balance the data.
Tensorflow allowed for custom metrics and for controlled
learning rate decays based on granular batch evaluation
scores. To analyze model performance, used
pyplot
to graph training and evaluation
metrics along with batch-level parameter trends similar
to Tensorboard graphing utilities.
Resulting \(R^2\) scores were poor and remained poor,
but improvements were observed throughout the research
term. To start, CNN models using transfer learning
resulted in
-0.16
. At present, improvements to model
parameters and dataset processing increased the score to
+0.1
.
Took initiative to begin development of a two-stage
model that classifies zero value observations against
non-zero value observations, then calculates a
regression accordingly. Currently in development.
Managed git versioning through GitHub as repository
owner. Responsible for writing and maintaining program
documentation with the intention of institutionalizing
knowledge gained from research.
Skills
Web Development
- Python
- Javascript
- HTML/CSS
- Node
- Dreamhost
Presentation
- Adobe Illustrator
- Whimsical
- Markdown
Data Science
- Tensorflow v2
- Tensorboard
- KerasTuner
- XGBoost
- CNN Models
- LaTeX
Data Analysis
- numpy
- pandas
- matplotlib
- pyplot
- statsmodels
References
Dr. Ciro Diaz
Data Scientist
Mai Nguyen
Accounting Analyst
Sunil Bansal
Product Owner
Niruja Nagarajah
HR Consultant