Seth Rice

Backend Engineer · Node.js, Python, LLM Orchestration
Elkhart, IN · Open to remote [email protected] r0ice.com/portfolio LinkedIn GitHub

Summary

Self-taught backend engineer running a 55-service production stack across a 2-VPS fleet — Twitch/Discord bot platform, multi-model LLM router, real-time community analytics, and paid SaaS products. ~270k LOC of Node.js, shipped and maintained alone. Currently completing a CS degree at Ivy Tech.

Skills

Languages Node.js, JavaScript, Python, Bash, SQL (SQLite)

Backend Express, REST APIs, WebSockets, Server-Sent Events, cron, systemd, PM2

APIs Twitch (Helix, EventSub, IRC/tmi.js), Discord.js, YouTube Data, Stripe, OAuth2

AI / LLM Multi-provider orchestration (Gemini, Groq, Mistral, Cerebras, Cohere, Anthropic, OpenAI-compat), prompt caching, parallel-fanout + tournament synthesis, RAG patterns

Infra Ubuntu VPS, Caddy reverse proxy, PM2 fleet, zero-downtime hot-reload (SIGUSR1), automated rolling backups, Discord-webhook alerting

Data JSON-at-rest, JSONL event logs, atomic file writes, SQLite migration design (230 MB → 100 MB)

Selected Projects

GameBot Platform — multi-tenant Twitch + Discord bot

2024 – present · Node.js, tmi.js, discord.js, PM2
  • ~270,000 LOC across 2,300+ JS files. Cog-based architecture, 100+ live feature modules, hot-pluggable via filename extension.
  • Runs 8+ isolated channel instances (per-tenant .env, per-tenant data directory, per-tenant Twitch/Discord credentials).
  • Built SIGUSR1 hot-reload for OAuth token rotation — eliminated PM2 restarts during 3-hour token-refresh cron (853 ms reload validated in prod).
  • Automated rolling backups with disk-aware skip (refuses to write when disk > 90% or free < 512 MB).

AI-Hub Masterbrain — 12-model LLM router

2026 · Node.js, multiple provider SDKs
  • Parallel-fanout LLM client routing prompts across Gemini, Groq, Mistral, Cerebras, Cohere, AI21, NVIDIA, Cloudflare Workers AI, HuggingFace, OpenRouter, Anthropic.
  • Tournament-synthesis: runs N providers in parallel, scores responses, returns synthesized best answer. Provider auto-disable on auth failure with 30-min cooldowns.
  • 5 downstream engines (autopilot, idea-engine, OKR, trend-injector, competitor-intel) unified through the same fallback chain — single point of routing logic.

Community-Live Intelligence — real-time presence + churn analytics

Node.js, Twitch Helix, JSONL, Server-Sent Events
  • Tracks 1,026,839 unique viewers across the platform.
  • 4-slice parallel scanner pages a 620k-channel pool against Twitch Helix in ~70 s using 15-account token rotation.
  • Multi-process pipeline: presence agent → churn sampler → leaderboard builder → HTTP API on port 8700 with SSE push to a live dashboard.

Sovereign Loop — autonomous build + staging pipeline

2026 · Node.js, MCP, Railway, OBS WebSocket
  • 24/7 self-pacing loop: Masterbrain proposes code changes → staging gate → human-approval webhook → live deploy.
  • Exposes a native MCP server (local stdio + bearer-gated HTTPS) so agents and the desktop client share one tool surface.
  • Integrated OBS scene/source control and Railway log access into the same agent loop.

vitals.r0ice.com & checkup.r0ice.com — health SaaS pair

2026 · Node.js, Express, JSONL, Caddy
  • Vitals: private biomarker + symptom + medication tracker. 36 biomarkers with optimal ranges, generates printable doctor-prep reports. Cookie auth, port 8803.
  • Checkup: public patient intake form for a clinic — 8 visit types, generates patient + provider reports with lab suggestions. Anthropic API slot wired.

Education

Ivy Tech Community College — Computer Science

In progress · Pell Grant

Planned transfer to Ivy Tech Bloomington campus.

Other