AI RESEARCH
2026-04-02
Topology of Thought
How curiosity, physics, and looking inside neural networks revealed something unexpected about how computation organizes itself
ONLINE
Adam Kruger
Systems Engineer | Homelab Enthusiast | AI Researcher | Autistic
99.8%
Uptime
Started my journey 13 years ago as a Cloud DevOps engineer. Over the years I've worked in security, project management, software engineering, and AI/ML engineering—now I kind of do it all. With one major exception: I'm doing everything in my power to push back against the cloud being the default for every task, no matter how large or small. Besides just being a nerd, this is why I got into homelabs.
Proxmox
Caddy
Python
Mojo
Go
FastAPI
Svelte
TailWindCSS
Temporal
Pydantic
MongoDB
Pgvector
System Resources
5/5 ONLINECPU
3%
RAM
33%
DISK
18%
NET
15.2 Mbps
7
Nodes
25
VMs / LXC
42
Services
268
vCPUs
575GB
Memory
51.0TB
Storage
$11180
Cloud Equiv/mo
Why Self-Host?
Cost Savings
This homelab replaces approximately $11500/month in cloud services. Over 3 years, that's $414,000 in potential savings after hardware costs.
Privacy & Control
All 23 VMs and their data remain on-premises. No cloud provider has access to databases, secrets, or application data.
Learning Value
Managing 7 hypervisors, 268 vCPUs, and various services provides hands-on experience with enterprise-grade infrastructure.
Flexibility
Full control over 575GB RAM and 52TB storage allocation. No vendor lock-in or arbitrary service limits.
AI Research Power
Run bleeding-edge ML research on consumer GPUs—no cloud A100s needed. Custom Mojo kernels achieve 12x speedups over PyTorch.
Security Research
Full control over model hidden states enables vulnerability research impossible on cloud APIs. Your data never leaves your network.
Equivalent Cloud Services:
EC2 + RDS + ElastiCache + EBS + CloudWatch
Virtual Machines
24 running| Name | Node | Type | CPU | RAM | Disk | |
|---|---|---|---|---|---|---|
| mattermost | homelab2 | vm | 2 | 4G | 30G | |
| metasploitable2 | homelab2 | vm | 1 | 1G | 8G | |
| OpnSense | homelab2 | vm | 4 | 4G | 64G | |
| attack-box | homelab2 | lxc | 2 | 2G | 16G | |
| vuln-docker | homelab2 | lxc | 4 | 4G | 31G | |
| kali-netdash | homelab3 | vm | 8 | 32G | 80G | |
| TimeScale | homelab1 | vm | 16 | 64G | 256G | |
| OMV | homelab1 | vm | 4 | 8G | 32G | |
| Prefect | homelab1 | vm | 8 | 8G | 96G | |
| RabbitMQ | homelab1 | vm | 8 | 8G | 96G | |
| SubSeek | homelab1 | vm | 4 | 16G | 32G | |
| Docker | homelab1 | vm | 16 | 8G | 128G | |
| forgejo | homelab1 | vm | 4 | 16G | 64G | |
| PyServices | homelab1 | vm | 8 | 8G | 96G | |
| DataServices | homelab1 | vm | 8 | 16G | 128G | |
| admin | mercury | vm | 4 | 8G | 64G | |
| site1 | mercury | vm | 4 | 16G | 64G | |
| redis-server | mercury | vm | 5 | 8G | 32G | |
| API-Gateway | mercury | vm | 4 | 16G | 64G | |
| Caddy-Router | mercury | vm | 4 | 4G | 32G | |
| Temporal | mercury | vm | 4 | 16G | 32G | |
| s3 | mercury | vm | 4 | 8G | 32G | |
| PowerDNS | mercury | vm | 2 | 4G | 32G | |
| workstation | local | workstation | 128 | 256G | 3600G | |
| pop-os-c2 | homelab4 | vm | 8 | 32G | 500G | |
| nexus | homelab4 | vm | 4 | 8G | 40G |
Running Services (11 containers + 31 systemd)
RUNNING
lab-stack-backend
lab-stack-backend
RUNNING
lab-stack-blog
lab-stack-blog
RUNNING
forgejo-runner
code.forgejo.org/forgejo/runner
RUNNING
cloudflared-blog
cloudflare/cloudflared
RUNNING
infisical
infisical/infisical
RUNNING
infisical-db
postgres
Recent Posts
AI RESEARCH
2026-03-23
Solving the Jane Street Dormant LLM Challenge: A Systematic Approach to Backdoor Discovery
How we solved all 3 backdoored DeepSeek V3 models using SVD weight analysis, persistent homology, multi-model deliberation, and 5,000+ indexed probes
HOMELAB
2026-01-28
12 Hours, 7 Layers, 1 Hard Wall: Debugging Memory on DGX Spark
A deep technical investigation into why MAX Engine allocates 117GB regardless of limits on NVIDIA DGX Spark's unified memory architecture - and why the solution doesn't exist in userspace