Programming

QTS supports the developer and trading-desk teams across RBC Capital Markets. The team needed to consolidate fragmented Git development platforms onto a single governed instance without disrupting active desks. I owned the automation layer end-to-end — design, build, rollout, and the reporting leadership used to make go/no-go calls on subsequent migration waves.

What I built

• Python automation orchestrating the migration of 300+ active projects between Git platforms, using REST APIs for inventory, repo transfer, and post-migration verification.
• Jenkins jobs that handled platform governance, enforcing access policies, tagging, and standardized CI hooks on migrated repos.
• Excel analytics that tracked migration progress and adoption metrics across business lines, used by QTS leadership to drive subsequent migration decisions.

FSRA regulates Ontario's non-securities financial sector, which includes pensions, credit unions, insurance, and mortgage brokering. As part of the Credit Unions team, I'm contributing to internal monitoring and risk modeling frameworks on the analyst side.

Current work

• Assisting in the build, data streaming, and validation of internal monitoring frameworks used by risk teams.
• Streamlining recurring internal processes through automation and improved documentation.
• Performing research feeding into the design and structure of internal risk reports.

The problem

Generative search engines (Perplexity, ChatGPT search, Google's AI Overviews) retrieve and cite content very differently from classical search crawlers. Most existing WordPress content is structured for traditional SEO and is not easily parsed by retrieval-augmented systems, which will rapidly gain in importance as generative search captures an increasing portion of the discovery layer.

What I'm building

• A modular, decoupled data pipeline that re-indexes WordPress sites for generative-search retrieval.
• A Python/PHP/Node.js multi-server architecture that organizes and classifies subpage schema into a structured, queryable format.
• A framework that generates JSON-LD from raw HTML using rule-based extraction combined with schema inference.

Why I find this interesting

It's a problem at the boundary of data pipelines, generative-search retrieval, and structured data enforcement. Generative engines are still in a stage where small structural changes to content meaningfully shift retrieval outcomes, which makes it a tractable engineering problem with real downstream value.

A Python analytics tool that pulls equity data via yfinance and evaluates instruments using risk-adjusted performance metrics, such as Sharpe ratio, volatility, and market-capitalization weighting. I extended it to model and visualize portfolio performance under different market regimes, which gave me a sandbox for thinking about hedging and trading implications without committing to a particular strategy framework.

The most useful thing the project did for me was to realize the discrepancy between the textbook version of modern portfolio theory and the actual return distributions once you pull real data - fat tails, regime shifts, and the way Sharpe ratios deteriorate in periods of market stress.

Ascribe, an Ottawa-based content agency that produces whitepapers, reports, and corporate announcements, wanted to understand how exposed their business was to generative writing tools. My mandate was both technical and strategic: build the thing they were worried about, then advise leadership on what it meant for their business.

What I built

• Hand-curated a training set of ~500 neutrally-toned articles relevant to business writing.
• A Python training loop that parsed each source article and produced a rewritten version in a target tone (professional, friendly, concise), generating a prompt-completion training file in the format OpenAI's fine-tuning API expected.
• Trained and evaluated a fine-tuned model that produced professional-tone rewrites measurably better than the base ChatGPT model of the time.

A browser game where the player runs a shop in a fantasy medieval market and their inventory, pricing, and outcomes are buffeted by a simulated world economy. I picked the medieval setting deliberately — it lets supply/demand, diminishing returns, and monopoly effects show up cleanly without having to model central banks or modern monetary policy.

It was my first JavaScript project of any real scope, and it taught me about both front-end development and the discipline of finishing a side-project and shipping incremental improvements.