Knowledge base

Everything I've learned building a sovereign AI stack, in public: the proof (case studies), the failure log (dead ends written up because they were the expensive part), and the interactive field guide.

πŸ“ The local-AI field guide

Interactive: what actually fits in your VRAM, what runs on each GPU tier, how many users one card serves, and when local beats cloud.

measured on an RTX 5090 Β· every number editable Β· open the field guide β†’

Case studies

Problem β†’ solution β†’ outcome, compressed to the result. Built and measured on my own stack: where a number is from a real run I say so, and where something didn't work I say that too. Each links to the full story.

I built a quantum ML lab. The honest answer: quantum mostly ties or loses today

Reproduced a published quantum advantage (0.900 vs 0.730), then erased it three ways: a fairly tuned classical model beat it, dequantization matched it, device noise took the rest. Eleven experiments, two real QPUs, zero manufactured wins.

PennyLane / Qiskit Β· 11 experiments vs classical baselines Β· ibm_fez + ionq forte-1 Β· barren plateau measured at 1,436Γ— full case study β†’ want this? email me β†’

One product photo β†’ a marketplace-ready 3D model + a compliance report

Five real SKUs on one RTX 5090: three passed clean, one refused, one flagged with the exact miss (703mm reconstructed vs 650mm listed). The gate stopping bad assets is the product.

TRELLIS.2 / Hunyuan3D Β· headless Blender Β· compliance gate Β· 63–154s per SKU Β· zero cloud full case study β†’ want this? email me β†’

One shot list β†’ an Unreal-ready previs pack in 15 minutes

Six props (4 photos + 2 text briefs): five delivered in 15m12s warm, one refused with its reason shown. One idempotent Unreal import script, Nanite on heroes only. Tested in Unreal Engine 5.8.

text/photo β†’ 3D Β· per-tier mesh chain Β· FBX + GLB Β· zero cloud full case study β†’ want this? email me β†’

A sovereign LLM that physically can’t leak

vLLM + Nemotron with routing-enforced privacy: the client profile refuses to fall back to cloud, failing loudly instead of degrading silently. Operated daily at up to 268 tok/s, 180K context.

vLLM Β· Nemotron Β· Tailscale-only Β· default-deny firewall the stack β†’ want this? email me β†’

Auditing a private codebase for vulnerabilities, fully air-gapped

Hybrid system beats the LLM alone, 0.97 vs 0.90, and the cross-file taint engine hits 1.00 precision where a per-file LLM gets 0.25. Honest headline: the fine-tune added ~0 detection capability, and measuring that caught 4 of my own eval bugs.

cross-file taint Β· Semgrep OSS Β· Qwen2.5-Coder QLoRA Β· runs under --network=none the auditor β†’ want this? email me β†’

A talking-head avatar from a photo and a voice sample, three quality tiers

For a 34.5s narration, measured: MuseTalk 78s, LatentSync 242s, Hallo2 1197s. Honest ceiling: social-grade; broadcast close-ups still go to a cloud path, and I say so up front.

Fish Speech clone Β· 3 lip-sync engines, isolated + pinned Β· zero cloud full write-up β†’ want this? email me β†’

Removing an object from production footage

Clean removal in 180 seconds, no ghost, background reconstructed. On a worst-case clip (glassware, low light) it leaves a residue, and clients hear that up front with a free sample frame.

VOID Β· SAM3 Β· SeedVR2 4K finish Β· failures shown full write-up β†’ want this? email me β†’

Starving a diffusion model of room to hallucinate

A vision LLM writes the scene constraints itself before any diffusion runs: 203 seconds end to end, unattended, 16 mask targets including 13 dependent reflections and shadows. The known hard case (a liquid) did not disappear, and that boundary is named.

Nemotron 3-pass forensic scan Β· SAM3 Β· VOID/VACE full write-up β†’ want this? email me β†’

Comic page β†’ narrated video, fully local

Panel detection, cloned character voices, narrator on captions, 2.5D parallax camera. Runs end to end; in active hardening on real series pages, hard problems documented as they’re solved.

vision LLM Β· OpenCV Β· Fish Speech Β· ffmpeg Β· zero cloud the project β†’ want this? email me β†’

β†’ What this costs and how to start

The failure log

New posts as the work happens. Looking for something specific? Browse by topic β†’ (LLM serving, GPU/VRAM, quantum, creative AI, audio, ops…)