Aurora, IL · CS @ UIC · open to DevOps + AI internships

Eugene Vincent

I build production infrastructure — and the AI that runs on it.

A CS student who runs a homelab like production: a k3s cluster with GitOps and full observability, plus a local + cloud LLM orchestrator that routes every task to the cheapest model that can handle it. Everything below is live from the rack under my desk.

Explore the work Get in touch
Live from the rack — k3s homelab updated every ~10 min Open dashboard
95.2% Server uptime
20 Minecraft tick rate · max 20
3 / 20 Players online
861 / 3072 JVM heap · MB
For engineering teams

I build and run production-grade infrastructure — k3s, GitOps, full observability, and cost-aware local + cloud AI — and I monitor it like it has paying customers.

For local businesses

I design and build fast, modern websites — like this one — that load instantly, meet accessibility standards, and cost almost nothing to host.

Selected work

Featured projects

Local AI · Python

Flagship

LLM Orchestrator

A working replication of NVIDIA's “Small Language Models are the Future of Agentic AI” — their six-step LLM→SLM conversion pipeline, run twice on my own agent telemetry. A converted 3.2 GB model beats a purpose-built 9.5 GB one 0.82 to 0.62 on a 61-task held-out set, and two different model families land on the same score — so the result belongs to the pipeline, not one lucky checkpoint. Every verdict came from a gate fixed before the run.

  • Python
  • SLM research replication
  • llama.cpp · 35B local
  • Claude API
  • evals + telemetry

Read the case study

Kubernetes · Observability

The Homelab

A single-node k3s cluster running Grafana, Prometheus, Loki and ArgoCD — GitOps-deployed, 95.2% uptime, with a Discord incident-analysis bot. The live metrics above come from it.

  • k3s
  • Prometheus
  • Grafana
  • ArgoCD

Read the case study

iOS · SwiftUI

Aria

A from-scratch SwiftUI music player — 10-band parametric EQ, 364 tests, zero third-party dependencies — backed by a Python/yt-dlp streaming service.

  • Swift 5
  • SwiftUI
  • AVAudioEngine
  • yt-dlp

Read the case study

See all projects

Toolbox

Skills

AI / LLMOps

  • Local inference — Qwen, quantized models
  • Tiered model routing (local + Claude)
  • Evals & cost-per-task tracking
  • ROCm / Vulkan · KV-cache tuning

DevOps / SRE

  • Kubernetes (k3s), Istio service mesh
  • GitOps with ArgoCD
  • Prometheus · Grafana · Loki
  • GitHub Actions CI/CD

Cloud / Platforms

  • Azure (Functions, AKS · AZ-900)
  • Cloudflare (Pages, DNS, Workers)
  • Docker · containerd
  • Tailscale mesh VPN

Languages

  • Python · FastAPI
  • Swift / SwiftUI
  • JavaScript / TypeScript
  • Vanilla web / PWA
Background

About

I'm Eugene Vincent, a third-year CS student at UIC who treats a homelab like production: version-controlled, monitored, and alerting me before my users — mostly friends on the Minecraft server — notice anything.

Today that rack runs a k3s cluster with GitOps deployments, a full Prometheus/Grafana/Loki observability stack, a Discord incident-analysis bot, and a local + cloud LLM orchestrator that routes each task to the cheapest model that can handle it. This site deploys to Cloudflare Pages but reads its live stats straight from the cluster.

I'm looking for DevOps and AI internships where reliability and cost-awareness matter — and I build clean, fast websites for local businesses on the side.

B.S. Computer Science, University of Illinois Chicago (expected 2028) · AZ-900: Microsoft Azure Fundamentals

Let's talk.

Hiring for a DevOps or AI internship, need a website that actually loads fast, or just want to swap homelab notes? I read everything and usually reply within a day or two.

Get in touch