Jake Schatzberg

Student network assistant, homelabber, and AI enthusiast

I am an Information Science student at Michigan State University, minoring in Information Technology, and I'm building a career in networking. At work I help keep MSU's campus network running, and at home I run my own homelab.

See what I've built | Get in touch

About Me

Jake Schatzberg

I have been interested in technology for as long as I can remember. One of my earliest memories is playing games on my grandma's laptop with her. Currently, I am a student network assistant at Michigan State University, where I troubleshoot network equipment across campus. Daily activities include troubleshooting switch configs, DNS and DHCP issues, creating and deleting DNS records, and fulfilling changes in the data center. I primarily work with Juniper and Infoblox equipment, and am JNCIA-Junos certified as well as Infoblox DDIP certified.

My work doesn't stop when I leave the office for the day. I make it a point to bring as many concepts as possible to my own home setup. At home, I am an avid Linux user, and run as much of my network as I can from my own systems. I have experimented with OPNsense as a firewall and router, BIND9 for DNS, Kea for DHCP, and tying them all together with NetBox for documentation and as a configuration front end. I also try to self-host as many services as I can through my UGREEN NAS running TrueNAS Scale, with plans to scale up to a Proxmox cluster with Kubernetes on top.

I have also been following AI since before the days of ChatGPT. In recent years, I have been most interested in open-source models that can run locally. This is an extension of my interest in homelabbing, where I want as little data leaving my own network as possible, while learning about the systems that the cloud companies have in place to train and serve their AI models.

Quick Facts

Skills

Networking

Systems

AI

Web

Projects

What I build in my free time.

Homelab

My homelab is where I host the services I use every day and try out tools I don't get to touch at work. It runs on a UGREEN NAS with TrueNAS Scale.

What it runs

  • Jellyfin: Movies and TV
  • *arr stack: Media management
  • Navidrome: Music streaming
  • SearXNG: Private search engine
  • NetBird: VPN for remote access, running through AWS
  • Immich: Camera roll
  • Nextcloud: Files
  • Forgejo: Git hosting
  • Coolify: Netlify alternative
  • Kavita: Books
  • Vaultwarden: Password management

Also experimenting with

  • OPNsense: Firewall and router
  • BIND: DNS
  • Kea: DHCP
  • NetBox: Network documentation
  • Proxmox: Virtual environments
What I learned

How to keep my data private and manage an internal network.

KiriKiri Z / KAGEX English Translation

The English visual novel scene is dominated by Ren'Py, and I think developers deserve more options. KiriKiri is the second most popular visual novel engine, yet its English documentation was practically nonexistent. I'm currently working on translating KiriKiri Z and the KAGEX engine that runs on top of it into English: the source code comments, the documentation, and the tools' interfaces.

How

  • Machine translation of the Japanese source, docs, and interfaces
What I learned

I now have a deeper understanding of the engine's architecture and the work needed to develop a visual novel in KiriKiri, as well as how to reverse engineer games made in it.

KiriKiri Z translation project on GitHub

Training Local AI Models

I've trained my own image and voice models and run them locally on an NVIDIA RTX 4070 Super. I have trained voice changer models with RVC, TTS with GPT-SoVITS, and image LoRAs for Stable Diffusion, Flux, and Anima.

Tools

  • kohya_ss
  • SimpleTuner
  • TensorBoard
  • RVC
  • GPT-SoVITS
What I learned

What parameters go into training AI models, such as steps, epochs, how to evaluate a training loss graph, and how to select an optimizer and learning rate.

Get in Touch

I graduate in and I'm looking for full-time roles in networking.