BA Computer Science from UNC Charlotte, Azure AI certified. I build production RAG systems, LLM-powered tools, and agent infrastructure — from safety harnesses that block dangerous AI actions in under 5ms to fintech products pulling live SEC data.
Currently focused on AI agent safety, MCP tooling, knowledge graph pipelines, and mechanistic interpretability research. Strong at explaining technical concepts to non-technical stakeholders.
Built AI prototypes from concept to deployment. Partnered with Product, Finance, and Operations to identify automation opportunities. Delivered demos and training to non-technical stakeholders.
Translate technical requirements for non-technical clients. Maintain cross-functional relationships across teams.
Sits between the AI agent and your codebase. Intercepts every file write, edit, and shell command — blocks dangerous ones in under 5ms. Runs Semgrep, secret detection, and tests at commit time. Seven-layer closed loop with LLMOps prompt optimization. Open source, zero paid dependencies.
Research project probing whether a model's residual-stream activations can predict chain-of-thought correctness. Found that a single direction in the prefill residual stream predicts answer correctness better than final-token activations after reasoning. Selective prediction at 50% coverage yields 71.6% accuracy vs 48.6% unconditional.
AI agent system with 20 MCP tools, YAML-defined pipelines, and human-in-the-loop review — swipe right to approve AI-generated code improvements, left to reject. React Native with 15 Zustand stores, Socket.io real-time updates, and 68 Maestro E2E flows.
Discord bot scrapes links from an entire guild, categorizes them with Claude, embeds them into a Neo4j knowledge graph (9 node types, 384-dim vectors), and exposes 8 MCP tools for RAG-powered Q&A from any Claude Code session.
Live fintech tool integrating 3 external APIs (SEC EDGAR, Polygon.io, FRED) into a unified valuation calculator. Implements Michael Burry's Owner's Earnings methodology — compares traditional Gordon Growth Model against a dilution-aware model that accounts for the real cash cost of stock-based compensation.
Open to collaborations, interesting problems, and conversations about AI tooling, developer infrastructure, or fintech.