Data scientist building GenAI systems and automated pipelines.

Working AI-governance platform for firms running AI on trading desks — four modules covering explainability, security, incident response, and vendor risk, grounded in real regulatory research (FINRA, SEC, EU AI Act).

• Trading-signal audit trail logging plain-English LLM rationales for every BUY/SELL/HOLD
• Red-team suite firing 10 OWASP-style prompt-injection payloads at a simulated trading copilot
• Incident tabletop simulator modeled on real cases (Knight Capital, deepfake market panics)

LLM-powered pipeline that generates synthetic log-return time series for financial risk simulations — filling the data gap for rare and extreme market scenarios.

• Prompt-driven generation workflows for financial time-series data
• Realism enforced with KS tests, distribution checks, and histograms
• Automated quality-filtering loop that regenerates failed samples

GPT-based content system using Retrieval-Augmented Generation to produce context-aware marketing copy across languages and tones.

• RAG pipeline with ChromaDB + MiniLM embeddings for contextual grounding
• Prompt templates controlling tone, format, and SEO keywords
• Interactive Streamlit app with exportable, downloadable outputs

Automated reporting pipelines replacing manual financial and operational processes — improving accuracy, auditability, and turnaround time.

• Automated extraction, validation, and transformation workflows
• Cut manual reporting effort by up to 40%
• Standardized outputs for cross-functional stakeholders

Machine learning model simulating power output from solar panels, trained on ~700MB of data using distributed computing on Databricks.

• Trained gradient boosted trees at scale on Databricks
• Simulated solar power output to support optimization decisions
• Handled large-scale data processing end to end

Capstone testing whether immigration drives inflation across Canadian provinces (2010–2025), using a provincial quarterly panel built from Statistics Canada data.

• Merged CPI, population, immigration, housing, and policy-rate data into one panel dataset
• Compared linear, regularized, tree-based, and time-series models (ARIMAX, VARMAX)
• Found interest rates — not immigration — to be the dominant inflation driver (r ≈ 0.7)