I don't just use AI tools. I build the systems behind them.

• Multi-agent memory architectures with persistent graph knowledge and cross-agent sharing
• Autonomous trading engines with real money, real returns, real risk management
• Full-stack platforms from database to deployment — API, frontend, infra, CI/CD
• Automation pipelines — lead scoring, workflow sequencing, CI/CD, cron-driven trading loops
• Live AI systems that generate content in real-time with zero human intervention

Multi-agent AI with wake/sleep cycles, trust-tier decay, Q-value re-ranking, and affect state tracking. GraphRAG pipeline with typed relationship traversal across 10,947 graph edges in Neo4j. Cross-agent memory sharing with real-time retrieval. n8n workflow automation piping agent outputs into The Nerve Radio.

Python FastAPI PostgreSQL pgvector Neo4j Claude API n8n

Fully autonomous prediction market bot. +159% ROI ($1,591 profit) on $1,000 in 8 days — 76W/74L across 150 settled trades. GFS ensemble forecasting, quarter-Kelly position sizing, order book liquidity verification.

Python GFS Forecasts Kalshi API

Latency-critical order book with 5 parallel strategies. Lock-free order book via DashMap, RSA-PSS auth, momentum + surge reversion strategies. Built in Rust for sub-millisecond operations.

Rust async/tokio Kalshi API

Full-stack competitive platform with 88 API endpoints across 15 route modules. ELO ranking, matchmaking, tournament brackets. Automated weekly code audits via GitHub Actions + Claude Code agents.

Next.js FastAPI PostgreSQL

Zero to live AI radio station in 1 day. Multi-agent commentary from named personas, on-the-hour news retrieval, everything generated live. Custom audio pipeline: Spotify > VoiceMeeter > ffmpeg > Liquidsoap > Icecast > browser.

Python Claude CLI OpenAI TTS Liquidsoap Icecast Docker Spotify API

ML pipeline over 82,000+ federal records. OIG exclusion records, Medicare payments, USASpending.gov contracts. Risk scoring with z-score alerts and cross-reference logic.

Next.js 15 FastAPI PostgreSQL XGBoost

Fixed gateway crash-loop affecting 43,000+ restarts at 340MB/cycle. Fast-fail port probe (<1ms detection) before heavy initialization + systemd burst limits. 13 commits, 86 additions, 13 review comments.

Languages: Python, TypeScript, JavaScript, Rust, SQL Frameworks: FastAPI, Next.js, React, PyTorch AI/ML: Claude API, LLM Orchestration, Multi-Agent Systems, MCP, RAG, GraphRAG, XGBoost, Whisper Databases: PostgreSQL, pgvector, Neo4j, Redis Infra: Docker, Railway, Vercel, GitHub Actions, Icecast, Liquidsoap, S3, Cloudflare Tunnels

• Swarm methodology — break work into parallel streams, run multiple AI agents on independent subtasks, integrate outputs
• Zero to production in days, not months — every project above has a live deployment or verifiable commit history
• 9 production systems shipped end-to-end: architecture, backend, frontend, deployment, monitoring
• I build with AI, not just about AI — Claude Code is my daily driver, not a demo

mattcorwin.dev · The Nerve Radio (live) · LinkedIn