Full Stack Engineer | AI Systems

Building production applications and intelligent systems with LLMs, RAG, and modern web technologies

Portfolio • LinkedIn • Email

MS in Computer Science @ Indiana University Bloomington | GPA: 3.9/4.0

• 🌐 Building production full-stack applications with real-time features
• 🤖 Developing multi-agent AI systems and intelligent LLM architectures
• 📊 Deploying RAG pipelines and MLOps infrastructure at scale

⭐ 1,500+ active users | 🏆 27 tournaments hosted

Full-stack platform built in collaboration with Keith Hedges. Real-time match synchronization, bracket management, and AI-powered match analysis. Built with modern monorepo architecture handling complex tournament state management and WebSocket communications.

Multi-agent AI system for autonomous code generation. Four specialized agents (Planner, Coder, Reviewer, Explorer) orchestrate complex coding tasks with sandboxed execution using E2B.

Production MLOps platform for real-time fraud detection achieving sub-100ms inference latency. Includes automated model retraining with drift detection and comprehensive Grafana monitoring.

Intelligent LLM router with semantic caching achieving 60% cost reduction. Routes queries to optimal models (GPT-4/GPT-3.5) based on complexity classification with 97% accuracy.

Jan 2026 - Present | Remote

• Architected AI-powered ETL platform using OpenAI GPT-4o for automated data pipeline generation, enabling natural language to SQL transformation across PostgreSQL, MySQL, and SQL Server
• Optimized LLM cost infrastructure through FinOps analysis, identifying 54.6% cost reduction via model downgrading, prompt optimization, and semantic caching strategies
• Developed AI field mapping service with LangChain and GPT-4o, automatically matching source to target columns with confidence scoring and 40% token reduction through metadata filtering

Dec 2025 - Present | Indianapolis, IN (Remote)

• Built real-time tennis match analysis system using Llama 3.3-70B via Venice.ai API, generating professional head-to-head predictions with streaming responses
• Developed web scraping infrastructure using Playwright headless browser with anti-detection measures, extracting H2H stats from matchstat.com
• Implemented TTL caching layer with thread-safe operations and automatic eviction, reducing redundant scraping by caching H2H data for 2 hours
• Deployed FastAPI backend with Server-Sent Events (SSE) for real-time streaming, Docker containerization, and Caddy reverse proxy

Jun 2025 - Dec 2025 | Hampton, IL (Remote)

• Architected production RAG system with 5-stage pipeline: query routing, reformulation, hybrid retrieval (BM25 + semantic), cross-encoder reranking, and GPT-4 generation, reducing document research time by 60%
• Built hybrid search engine combining Sentence-BERT embeddings with BM25 using Reciprocal Rank Fusion (RRF), achieving 94% retrieval relevance on 10,000+ environmental documents
• Developed LLM-powered data extraction pipeline using GPT-4 function calling, achieving 95% accuracy and reducing manual extraction from 3 hours to 15 minutes per document