Ciena

• Implemented pipelines to ingest raw customer data from streaming and batch inputs and assembled preprocessed datasets used for training anomaly detectors used in production
• Aggregated and visualized customer devices’ alarm occurrence data to determine the types of facilities that have been shown to experience specific alarms, the occurrence frequency of those alarms, and the feature distributions associated with each alarm
• Researched and cross-validated different multivariate time series classification and forecasting frameworks to inform and justify the use of inference models in production

Stack

• Python
• Spark
• Kafka

Apption Software

• Built a server to query OpenStreetMap data stores against Canada Post databases resulting in visualizations and statistical analyses of OSM coverage of Canadian provinces
• Developed a Tableau workbook (.twb) English – French translation tool producing fully localized data visualizations and dashboards used in executive overviews of top 12 clients’ reports
• Implemented a detection algorithm using approximate string matching techniques to extract unique move-in candidates from postal address and occupancy databases

Stack

• Python
• Tableau

Systopia

Vertex and Edge ordering for optimizing graph processing. Prior work showed that reordering the vertices or edges of the graph speeds up computation of workloads like PageRank and Connected Components. In my Masters research, I:
• Evaluated the combined effect of a group of vertex and edge ordering techniques on a dataset of large, in-memory, real-world graphs (e.g. social, hyperlink, road networks).
• Parallelized the SlashBurn vertex reordering algorithm using Afforest.
• Proposed a novel, lock-free, multithreaded vertex-and-edge ordering technique that leverages the compressed SlashBurn graph isomorphism and traverses the edges of the graph using the Hilbert Space Filling Curve.

Stack

• C++
• OpenMP
• Python

Computational Geometry Lab

• Developed a graphical model of supply chain structures to facilitate the automatic detection of underutilized resources in production processes
• Implemented heuristic contraction and partitioning procedures based on structures present in supply chain graphs to reduce dataset size and complexity and improve performance of a network-separation algorithm
• Tested runtime and correctness of community detection algorithms (Leiden, Louvain) to evaluate the feasibility of use on customer product structures

Stack

• Python
• Gephi
• igraph