I'm a -year engineering veteran known for delivering large-scale, mission-critical systems and creating opportunity through technical initiative. My expertise spans distributed systems, big-data platforms, and high-reliability infrastructure. Recently, I've focused on applied AI—integrating models into real products, optimizing performance, and enabling teams to ship fast with confidence. I'm drawn to roles where deep systems engineering and modern AI innovation meet to drive outsized impact.
Co-developed an E2E model onboarding platform (patent pending) for RAI models. Co-authored an ML model/paper for steganographic image watermarking. Optimized in-house image classification and code vulnerability models, cutting >$10M/year in COGs and reducing P99 latency 3×.
Created tooling & insights over Petabyte-scale telemetry data through a C# domain-specific-language.
Refined an ML testing platform for reliability, scalability, and performance on Azure.
Co-developed the organization's first HPC cluster for Ground Truth Analysis on Azure.
Developed test infrastructure for third party developer APIs used by the Windows 8 Calendar app.
Owned the test infrastructure for the entire sync stack across all Windows Live Client Applications.
Designed PCBs and programmed firmware for a suite of home automation products.
Tested releases of internal payroll software, replacing the aging COBOL solution.
| Type | Title | Details | Products |
|---|---|---|---|
| Publication | InvisMark: Invisible and Robust Watermarking for AI-generated Image Provenance | arXiv 2411.07795, Co-author (2024). Invisible, tamper-resistant watermarking for AI images; preserves quality (PSNR ≈ 51, SSIM ≈ 0.998) and robustly decodes UUIDs with >97% accuracy. |
Bing Image Creator Microsoft Designer Microsoft Copilot AOAI Image Models: - DALLE 3 - GPT-Image-1 - Sora |
| Patent | Knowledge-Driven Pre-Trained Form Key Mapping | US12423949B2 - (2025) Automatically maps form fields to standardized keys using a language model, improving accuracy and scalability in document processing. | Microsoft Form Recognizer |
| Patent | Training and Applying a Key Sentence Classifier Model | US20250356123A1 - (2025) Retains only output-relevant sentences in LLMs, reducing compute and memory use while speeding responses. | Internal Model Training |
| Patent | Customer-Driven Responsible AI Model Update Framework | Filed 2025. Uses customer dissatisfaction reports to generate targeted training data for faster, cheaper, and more accurate AI model updates. | All Microsoft AI Products |
| Patent | User-Defined Topic Classification and Moderation | Filed 2025. No-code system for custom topic classification with pruned LLMs; enables scalable, real-time filtering with minimal data. | Azure AI Content Safety |
| Patent | Intellectual Property Risk Detection in Code Snippets – Evaluation and Iteration | Filed 2025. Detects IP risks in AI-generated code using snippet fingerprinting and adaptive filtering, reducing false positives and improving safety. | GitHub Copilot |
During my tenure at RIT, they were ranked #1 in the nation for their CE program. I graduated one trimester early, with a 3.61 GPA.
School Mascot: The M&Ms
