Compute Infrastructure

AI Studio Architecture

Two-system architecture separating production AI services from design and development workloads. All machines connected via 10Gb backbone with SSH key-based orchestration.

NIK

Primary AI Server — Always-On 24/7

• i9-10900K · 32 GB DDR4 · RTX 2080 Super 8GB
• Ubuntu 24.04 LTS · Headless
• Open WebUI · AnythingLLM · n8n (Docker)
• Faster-Whisper STT (CUDA int8)
• Storage: 2.3 TB (NVMe + LVM)
• Ollama: 6 models including custom NIK persona

Brain — RAG + Orchestration

ECHO

Data Collection & Secondary AI

• i9-10980HK · 32 GB DDR4 · RTX 3080 16GB
• Linux Mint 22.3 · Kernel 6.17
• TeslaQuake data collectors deployed
• NOAA SWPC + INTERMAGNET processing
• Storage: 2.8 TB across 3× NVMe
• Battery UPS — surge-protected, sleep disabled

24/7 Data Gathering

FORGE

GPU Powerhouse — AI Vision & Rendering

• i9-13900KS · 64 GB DDR5 · RTX 4090 24GB
• Windows 11 Pro · CUDA 13.2 · 10Gb SFP+
• IRIS: custom Gemma4:31b vision model
• Ollama: 7 models (181 GB total)
• Llama3.2-vision:90b, Llama3.3:70b
• Swappable between System 1 and System 2

22.5 GB VRAM Available

STUDIO (R16)

Command Center & Daily Operations

• i9-14900KF · 64 GB DDR5 · RTX 4090 24GB
• Windows 11 Pro · 3-monitor setup
• Claude Code + Claude Desktop
• TeslaCoder v3 dashboard host
• SSH keys to all machines (silent auth)

Orchestration Hub

LAB (R13)

Design Development — FreeCAD + Blender + IRIS Vision

• i9-12900KF · 32 GB DDR5 · RTX 3080 Ti 12GB · Linux Mint 22.3
• FreeCAD (2D permits + 3D design) · Blender 4.0 (photorealistic renders) · Direct 10Gb DAC link to Forge for heavy render offload

System 2 — Design Pipeline

Geophysical Data Archive

Data Inventory — 13 Monitoring Categories

563 GB of geophysical data collected across 80,000+ files, distributed between NIK and ECHO machines. Spanning solar, geomagnetic, Schumann resonance, telluric, ionospheric, seismic, atmospheric, bioacoustic, and more — with some datasets reaching back to 1934.

Category	Research Factor	Total Size	Files	Coverage	Gap Status
Solar / Geomagnetic	W	58.1 GB	20,085	1963–2026	No gaps in analysis window
Schumann Resonance	R	35.3 GB	36,091	2000–2026	20 gap years (1980–1999)
Telluric	T	28.2 GB	5,941	2005–2026	31 gap years
Ionospheric	I	215.5 GB	22,916	1950–2026	No gaps in analysis window
Celestial	Z	197.5 MB	461	1900–2026	40 gap years
Seismic	S	1.2 GB	3,547	1990–2026	No gaps in analysis window
Atmospheric	A	140.8 GB	5,178	1934–2026	No gaps in analysis window
Bioacoustic	B	75.8 GB	10,289	2015–2025	37 gap years
Animal Behavior	AN	7.6 GB	1,716	In collection	Emerging category
Ocean / Tidal	O	400.5 MB	1,175	1909–2025	35 gap years
Volcanic	V	6.0 MB	42	2026	46 gap years
Gravity	G	118.1 MB	21	In collection	Emerging category
Events / GDACS	E	1.1 GB	457	2010–2026	30 gap years
Total		563 GB	107,919

Signal Processing

Triple-Algorithm Anomaly Detection

Three complementary statistical algorithms running in parallel, each detecting different anomaly signatures. Combined output feeds a proprietary accumulation model protected under pending patent.

Welford

Online streaming algorithm computing mean and variance per metric, weekday, and month. Detects statistical spikes with z-score thresholds at 1.5 (low), 2.0 (medium), 3.0 (high), and 4.0 (critical). Requires minimum 10 samples before activation.

CUSUM

Two-sided cumulative sum for regime shift detection. Identifies sustained directional drift in metrics that may not trigger spike-based detectors. Complements Welford by detecting gradual rather than sudden changes.

EWMA

Exponentially weighted moving average with α=0.2 balancing reactivity and stability. Flags when smoothed trends breach control limits or sustained directional movement is detected. Provides trend context for the other two algorithms.

Data Management

Database Architecture

108 PostgreSQL tables in Supabase managing real-time data ingestion, historical archives, anomaly detection state, and correlation analysis — migrating to fully local infrastructure.

• Active Tables: 108 in public schema
• Largest Tables: SuperMAG stations (943K rows), IERS polar motion (23K rows), Schumann readings (359K rows)
• Real-Time Feeds: NOAA Kp/Dst, GOES X-ray, solar wind plasma/mag, SWPC alerts (169K rows)
• Historical Archives: 358K historical earthquakes (25 years M4.0+), 76K daily sunspot numbers (1818–present)
• Anomaly State: 6,804 detected anomalies with z-scores, 16 CUSUM state records, 16 EWMA state records
• Migration: Transitioning from Supabase cloud to local PostgreSQL on NIK/ECHO

Institutional Data Sources

Source Provenance

All data sourced from established governmental, academic, and research institutions.

• USGS — Earthquake Hazards Program
• NOAA SWPC — Space Weather Prediction Center
• NASA OMNI — Merged solar wind data
• NASA CDDIS — IONEX ionospheric data
• NASA SDO — Solar Dynamics Observatory
• NASA DONKI — Space weather events
• GFZ Potsdam — Official Kp Index (1932–present)
• Kyoto WDC — Dst Index (1957–present)
• SILSO/ROB — Sunspot Number (1700–present)
• INTERMAGNET — Global magnetometer network
• SuperMAG — 943K+ station records
• HeartMath GCI — Schumann resonance monitoring
• Tomsk SOS-70 — Schumann observatory
• Cumiana VLF — VLF monitoring station
• IERS — Earth orientation parameters
• GDACS — Global disaster alerts

What Research Funding Enables

Research Objectives

The infrastructure is built. The data is flowing. These are the next steps that require institutional support, independent oversight, and peer collaboration.

Independent Peer Review

Submit correlation methodology and observation data to independent seismological and space weather researchers for formal peer review and publication in a refereed journal.

Priority

Blind Validation Study

Design and execute a prospective blind validation study with pre-registered methodology, time-stamped observations, and independent oversight to establish statistical significance.

Priority

Expanded Sensor Network

Deploy additional Schumann resonance and telluric monitoring stations to fill data gaps — particularly the 20-year Schumann gap (1980–1999) and 31-year telluric gap.

Infrastructure

Historical Backtesting

Complete formal backtesting across the 1980–2026 analysis window using the four gap-free categories (Solar, Ionospheric, Seismic, Atmospheric) against the full USGS M4.0+ historical record of 358,000+ events.

Validation