Hi I’m Jeremiah.
Metafizik is the container for the worlds I build.
My background spans data science, AI, blockchain, music, animation, and real-time platforms — but the through-line is always the same:
turning ideas into places and platforms that people can explore.
Currently operating between physical and digital dimensions.
GitHub: https://github.com/Metafiziks/
LinkedIn: https://www.linkedin.com/in/jeremiahwilliams1/
Email: jeremiah@metafizik.io
Mesmerizing the masses with meticulously made movements
How the Metafizik Deployment Architecture works
This diagram shows the full delivery loop for Metafizik projects: ideas start as GitHub issues, Copilot Agent executes against the repo, development runs in Codespaces, GitHub Actions validates and deploys, and changes are shipped to both Azure and Google Cloud. Production telemetry then feeds directly back into the next development cycle.
GitHub Issues & Planning
Work begins as structured issues, roadmap items, bug reports, and implementation tasks.
Copilot Agent + Claude
The coding agent interprets issues, edits code, writes migrations, and prepares PR-ready changes.
Codespaces Dev Environment
Development happens in a ready-to-run cloud workspace with repo context, tools, and isolated branches.
GitHub Actions CI/CD
Automated workflows validate code, authenticate to cloud services, apply migrations, and trigger deployments.
Multi-Cloud Deployment Pipeline
Infrastructure and application updates deploy across both Azure and Google Cloud using automated workflows.
Running Services (Azure + GCP)
The deployed system runs across Azure and Google Cloud, handling live traffic and production workloads.
Telemetry & Observability
Every deployment and user interaction produces telemetry for debugging, health monitoring, and system learning.
Feedback Into Delivery
Production behavior feeds back into new GitHub issues, which the agent turns into the next round of changes.
Agent Delivery Pattern
Issue is written in plain language → coding agent implements the change → workflow validates it → Azure and Google Cloud deploy it → telemetry confirms whether it worked.
Why Codespaces Matters
Codespaces keeps the environment consistent across debugging sessions, PR work, and agent-generated changes, reducing local-machine drift.
Why GitHub Actions Matters
Actions becomes the control plane: tests, cloud auth, schema migrations, and deployment all run through one repeatable automation layer.
What GitHub does
GitHub is the coordination layer. It holds the backlog, codebase, branches, pull requests, and workflow automation triggers.
What the agent does
The agent translates issues into implementation work: code edits, workflow fixes, schema changes, and debugging instrumentation.
What the cloud layer does
Azure and Google Cloud together form the runtime and deployment layer, hosting services, scaling workloads, and executing production logic.
What the loop means
This is not just CI/CD. It is an agent-assisted delivery system where production telemetry feeds directly back into the next development cycle.
Projects
From idea, to conception, to perfection.
GRIZL AI Support Agent
grizl.xyz/chat
An AI-powered knowledge and interaction system, combining retrieval-augmented generation (RAG), hybrid search, self-learning feedback loops and multi-agent orchestration.
Tech: Microsoft Foundry, Azure OpenAI, Azure AI Search, Node.js, Express, React, GitHub Actions, Azure Container Apps, Google Cloud Run, BigQuery, Workload Identity Federation, GitHub Copilot Agent, Codespaces
GRIZL Self-Healing Support OS
A digital immune system built on Azure — detecting, triaging, and remediating issues through governed AI, automated observability, and approval-gated remediation.
User → GRIZL Chat
- Chat widget embedded in Squarespace
- POST /api/chat with user message
- Query rewriting before retrieval
- Session + request telemetry captured
Azure AI Processing
- Azure AI Search — hybrid + semantic retrieval
- Azure OpenAI — grounded chat completions
- App Insights — event and trace ingestion
- Log Analytics — centralized diagnostics
Telemetry + Detection
- App Insights event stream
- KQL anomaly detection queries
- Severity classification per signal
- Anomaly triggers downstream triage
Analytics & Dashboards
- Application Insights workbooks
- Log Analytics KQL queries
- Power BI system health reporting
- Real-time remediation visibility
Anomaly Engine
- Latency, fallback, and quality rules
- Deduplication + cooldown windows
- Severity scoring and routing
- Anomaly payload emitted to triage
Foundry Triage Agent
- Deterministic, temperature-0 triage
- Grounded decisions via runbook context
- actionType + decisionTrace output
- App Insights trace on every decision
Workflow Orchestration
- anomaly → triage → grounding → issue update
- executionHint routes next action
- GitHub issue enrichment
- Policy + runbook grounding per step
Supervised Remediation
- Approval-gated draft PR creation
- Post-remediation verification run
- Remediation memory + outcome logging
- Adaptive confidence from history
What Azure does
Provides retrieval, generation, and observability in one platform — AI Search, OpenAI, App Insights, Log Analytics, and Foundry runtime all live here.
What GRIZL does
GRIZL is the live runtime and primary signal source. Every chat request, retrieval attempt, and model response becomes telemetry that feeds the detection layer.
What telemetry does
Acts as the system's nervous system — streaming events into App Insights, running KQL anomaly rules, and firing triggers that wake the triage pipeline.
What Foundry + GitHub do
Foundry handles governed, grounded reasoning. GitHub is the coordinated execution surface — issues, PRs, and Actions deliver remediation under human approval.
GRIZL grizl.xyz
An AI-enhanced social economy where digital companions remember, evolve, and perform. Chat with characters shaped by memory, simulated emotion, and battle history. Every interaction becomes part of an ongoing narrative.
Tech: Node, React, PostgreSQL, MongoDB Atlas (Vector Search), LangChain, GPT-4o / Claude, AWS, GCP, Vercel, Hardhat Ethereum, Kling, Seedance 1.5 Pro, ElevenLabs, Stripe, GoogleAuth.
Where Digital Avatars Battle, Trash-Talk,
and Build Empires
Challenge an NFT
Pick your NFT, choose a target from any collection. Set tier (Casual → All-In) and stance. GRFTY wager escrowed.
Defender Accepts
Defender picks their stance and locks in. GRFTY escrowed. 15-min betting window opens for spectators.
Best of 5 Rounds
Each round weighs 5 traits differently. Momentum, pressure, clutch mechanics. AI narrates round-by-round.
Winner Takes All
Winner gets wager back + prize + XP. Battle summary card with full breakdown appears in feed.
⚡ Battle Tiers
| Tier | Wager | Win | XP |
|---|---|---|---|
| CASUAL | 10 | +10 | +30 |
| STANDARD | 25 | +25 | +50 |
| HIGH STAKES | 50 | +50 | +75 |
| ALL-IN | 100 | +100 | +100 |
🎯 Round Trait Weights
💰 GRFTY Token
- Start: 100 GRFTY per user
- Earn: Win battles, social engagement, daily activity
- Spend: Battle wagers, GRFT Packs, marketplace trades
- Escrow: Wagers locked at challenge/accept, settled after battle
🃏 GRFT Packs
- Unlock at: 1,000+ lifetime GRFTY earned
- Cost: 100 GRFTY per pack • Contains: 3 random memes
- Supply: Seasonal limited drops
- Tradeable: Sealed packs on marketplace
- Memes degrade: 5 uses then burned
- NFTs attach 1 meme/post/24hrs for XP boost
🏪 Internal GRFTY Marketplace
- Buy Now: Instant purchase at listed price
- Auctions: Bid on rare memes & sealed packs
- Zero rake — No platform fees on trades
- 100% GRFTY denominated — closed-loop economy
- Meme condition matters: 5/5 uses worth more than 1/5
SoniQute soniqute.com
SoniQute is a full-stack music discovery ecosystem built from scratch — connecting emerging artists to fans through physical collectibles, AI-generated dance videos, and a social participation engine that turns every share into a real listen.
Tech: Next.js · Node.js / TypeScript · MongoDB · Seedance AI · Shotstack · Kling AI · Cloudinary · Ethereum / Hardhat · Google Cloud Run
Where music,
collectibles & dance
become one system.
SoniQute is a full-stack music discovery ecosystem built from scratch — connecting emerging artists to fans through physical collectibles, AI-generated dance videos, and a social participation engine that turns every share into a real listen.
artist catalog
QUTIE Packs
AI video render
real listen = real data
via SigniFi tracking
to your door
Music Discovery
Emerging artists distribute via DistroKid. Every dance video attaches the artist's track. Every TikTok post is a verified listen, ISRC-tracked across platforms via SigniFi.
Collectible Universe
Three interconnected physical collectible universes — PaMs, QUTIE PaMs, and QUTIE Chums — each with original lore, theme songs, and faction world-building. Physical card packs, 5 cards each, distributed rarity.
Dance Studio AI
Users pick a QUTIE Chum, choose dance vibes, and render a 15-second 1080p video. Seedance generates clips from static images. Shotstack BPM-stitches. Kling applies lipsync.
The Dance Studio AI Pipeline
A fully automated video generation system where every step flows into the next — from a static character image to a lipsync-ready 1080p video with music, transitions, and text overlays. Built entirely from scratch.
Scene Generation
User selects a vibe (Hype, Chill, Fierce…). Seedance 1.5 Pro generates a 5s 1080p clip from the character image.
~2 min · Wavespeed APIBPM Stitch
3 scenes stitched via Shotstack with beat-snapped cut points, user-selected transitions, and text overlays.
Shotstack · transitions: fade/zoom/wipe/+4Kling Lipsync
Stitched video uploaded to Cloudinary, passed to Kling AI for lipsync against the selected music track clip.
~6-7 min · Wavespeed KlingFinal Render
Output stored in Cloudinary qutie-final/, surfaced in the in-app Render Library. +5 SONIQ points awarded.
Cloudinary · qutie-final/