Audited an AI/data platform's brand visibility across major LLMs and found the gaps: no comparison or integration pages, and a competitor capturing citations the client should have owned. Published 13 pieces to fill those gaps, then re-measured on the same query harness. Citations roughly doubled across all models — comparison queries went from 1.0 to 2.0 citations per answer — and the gains held on a model that launched after the content was published, showing the results generalized rather than being overfit to one platform.

Designed and ran a series of controlled experiments — control and treatment groups with statistical significance testing — to isolate what drives brand citations in LLM responses. The headline finding: original research content outperformed standard content by 10–100x, and pairing it with a formal tone raised the probability of brand citation 7–15x on ChatGPT. Those findings became the backbone of the client's content strategy.

Audited the client's full content library against ~740 test queries across LLM platforms. The diagnosis: a visibility problem, not a competitiveness problem — when LLMs mentioned the brand, it beat competitors; it just wasn't surfacing often enough. Delivered specific content recommendations plus an interactive dashboard so their team can track AEO performance over time.

Developed and maintained fraud detection models for insurance clients, working in C# against client-specific data schemas. Included building data pipelines to process large extracts, loading them into SQL databases, and writing complex queries for feature engineering. Led client-facing meetings to communicate implementations and gather feedback.

Built for a lead-generation client, this pipeline brought together data from multiple advertising channels using Python and SQL, then surfaced performance insights across acquisition funnels. The output was a set of KPI dashboards tracking CAC, conversion rates, and campaign ROI — enabling stakeholders to self-serve their reporting rather than waiting on ad hoc analysis.

Developed and deployed a Streamlit dashboard that gave a healthcare client real-time visibility into conversion performance. Included interactive visualizations, parameter controls, and drill-down capability — built to be used directly by non-technical stakeholders without needing to ask a data team for reports.

Developed and executed quantitative trading strategies at a proprietary trading firm, tracking performance metrics and iterating based on data-driven analysis. Built a real-time market data pipeline in Python and SQL for continuous ingestion and querying of high-frequency trading data, and a backtesting module for evaluating strategies before deployment.

Built LLM-powered content generation tools in Python as part of a broader data platform for a B2B client. The tools were designed for production use, with managed SQL databases and a clean integration into stakeholder-facing workflows. Included prompt engineering, output quality checks, and iteration based on stakeholder feedback.