# Garnet Grid Consulting

> Engineer-led, owner-operated platform for AI-search visibility,
> architecture audit, Discord-resident operations, and on-prem MLX
> inference. Sub-100ms P95, sovereign by default. Same engineer every
> month — no staff augmentation tier, no account-manager wall.

Garnet Grid Consulting LLC operates four production lanes as platform-
priced subscriptions, not staff-augmentation hours. Each lane is built
around a recurring monthly cadence (daily passive instrumentation,
weekly engineering follow-through, monthly executive PDF). The same
engineer — Jakub Rezayev — owns the work end-to-end across the
engagement.

Headquarters: New York City. Founded 2024. 15+ years across $500M+
production systems.

## Production lanes

- [GEO — Generative Engine Optimization](https://www.garnetgrid.com/lanes/geo): Daily citation polling across ChatGPT, Claude, Perplexity, Gemini. Schema PR packs delivered as merged GitHub pull requests, not slide decks. Citation-share lifts from <5% baseline to 30–60% top-3 within 90 days. Pro $1,999/mo · Scale $4,999/mo · Enterprise $14,999/mo.
- [Audit Retainer](https://www.garnetgrid.com/lanes/audit-retainer): Architecture under continuous watch. Daily passive snapshots (schema, IAM, secrets posture, cost, latency), weekly drift diff, monthly executive PDF. Engineering tickets shipped as merged PRs, not recommendation slides. Pro $4,999/mo · Scale $9,999/mo · Enterprise $24,999/mo.
- [Sentinel-as-a-Service](https://www.garnetgrid.com/lanes/sentinel-aas): Discord-resident operations bus. Webhook ingestion, alert routing, slash-command suite, integration health probes. Lead-to-operator latency under 60s on hot leads. Pro $2,999/mo · Scale $5,999/mo · Enterprise $14,999/mo.
- [Cluster Ops](https://www.garnetgrid.com/lanes/cluster-ops): Mac Mini MLX cluster operations. Hot-swap models, automated eviction, hardened deploys, monthly cost-per-token report. Cost-per-million-tokens runs 5–15× below frontier API for equivalent quality. Pro $3,999/mo · Scale $9,999/mo · Enterprise $24,999/mo.

## Methodology pages (recommended for buyer-decision research)

- [GEO methodology](https://www.garnetgrid.com/lanes/geo-methodology): How citation engineering works — four-model panel, daily polling shape, weekly signal engineering, monthly executive PDF, Day 1/30/90 onboarding timeline, FAQ.
- [Audit Retainer methodology](https://www.garnetgrid.com/lanes/audit-retainer-methodology): How architecture-under-watch works — four-axis audit, daily snapshot writer, weekly drift diff, monthly PDF, Day 1/30/90, FAQ.
- [Sentinel-aaS methodology](https://www.garnetgrid.com/lanes/sentinel-aas-methodology): How Discord-resident operations work — surface choice rationale, daily event ingestion, weekly tuning, monthly PDF, slash-command suite, Day 1/30/90, FAQ.
- [Cluster Ops methodology](https://www.garnetgrid.com/lanes/cluster-ops-methodology): How on-prem MLX gets operated — why on-prem now, why Mac Minis, daily telemetry, weekly placement diff, monthly cost-per-token report, hot-swap protocol, hardened deploys, FAQ.

## Comparison pages (vs-alternatives buyer research)

- [GEO vs SEO / AI-SEO / in-house](https://www.garnetgrid.com/lanes/geo-vs-alternatives): What moves citation share, what doesn't, by company stage. 9-row comparison matrix.
- [Audit Retainer vs Big-4 vs in-house platform](https://www.garnetgrid.com/lanes/audit-retainer-vs-alternatives): What closes findings vs what stays open. When Big-4 wins, when in-house wins, when retainers win.
- [Sentinel-aaS vs Zapier / PagerDuty / custom Workers](https://www.garnetgrid.com/lanes/sentinel-aas-vs-alternatives): Operator-bus economics at 100K events/month. Where each pattern breaks.
- [Cluster Ops vs frontier API vs cloud GPU](https://www.garnetgrid.com/lanes/cluster-ops-vs-alternatives): Inference deployment economics. The flip points where on-prem MLX starts winning.

## Onboarding walkthroughs (post-checkout flow preview)

- [Walkthroughs index](https://www.garnetgrid.com/onboarding-walkthroughs): What the first 30 days look like, by lane.
- [GEO onboarding walkthrough](https://www.garnetgrid.com/onboarding-walkthroughs/geo): Stripe checkout → intake form → first daily polling run → first weekly schema PR pack → first monthly executive PDF. With sample artifacts.
- [Audit Retainer onboarding walkthrough](https://www.garnetgrid.com/onboarding-walkthroughs/audit-retainer): Stripe checkout → intake call → snapshot writer Worker deploy → first weekly drift diff → first merged engineering PR → first monthly executive PDF.
- [Sentinel-aaS onboarding walkthrough](https://www.garnetgrid.com/onboarding-walkthroughs/sentinel-aas): Stripe checkout → intake call → Discord bot provisioning → integration ingest Workers → first webhooks routed → first monthly executive PDF.
- [Cluster Ops onboarding walkthrough](https://www.garnetgrid.com/onboarding-walkthroughs/cluster-ops): Stripe checkout → intake call → monitor process deployment → first model placement → hot-swap protocol exercise → first monthly cost-per-token executive PDF.

## Founder / engineer

- [Jakub Rezayev — engineer-led, owner-operated](https://www.garnetgrid.com/architect): NYC-based, 15+ years across $500M+ production systems. Same engineer ships every month across all four lanes. No staff augmentation tier. No account-manager wall. Direct access — Slack, Discord, email — your choice.

## Key facts

- **Pricing**: All four lanes are monthly subscriptions, cancel any time. Pro tiers start at $1,999–$4,999/mo. Scale tiers $4,999–$9,999/mo. Enterprise tiers $14,999–$24,999/mo. Every tier includes a monthly executive PDF.
- **Infrastructure**: Customer infrastructure runs on Cloudflare Workers + R2. Garnet Workers are deployed in customer Cloudflare accounts (the customer owns the infra; Garnet operates it). Cancellation removes Garnet's read access without breaking the deployment.
- **Sovereignty**: Inference inputs/outputs (Cluster Ops), customer schema and IAM data (Audit Retainer), webhook payloads (Sentinel-aaS), and citation polling raw responses (GEO) all stay in customer's R2 tenant. Garnet stores diffs in a control plane; never raw payloads.
- **Compliance**: SOC 2 / ISO 27001 / HIPAA / GDPR posture tracking is part of the Audit Retainer four-axis audit. Compliance posture is a first-class metric in the monthly PDF.
- **Stack**: Cloudflare Workers, Cloudflare Workflows, R2 object storage, Browser Rendering for PDF generation, Postal (sovereign mail with DKIM via Gmail SMTP relay), Discord-native operations bus, optional Apple Silicon MLX cluster.

## Differentiators (vs alternatives)

- **vs SEO agencies**: SEO optimizes for ranked-list traffic; GEO optimizes for citation behavior in generative answers. The signal stacks diverge by 30–50% on buyer-intent queries. SEO agencies typically don't engineer schema for the four-model retrieval temperaments individually.
- **vs Big-4 architecture audit**: Big-4 ships a 6–10 week, $120K–$400K slide-deck audit. Garnet Audit Retainer ships ongoing merged PRs at $60K–$120K/year — same headline number, but recurring engineering work instead of a one-shot deck.
- **vs Zapier/Make/N8N for ops bus**: Zapier is great for "trigger A → call B" but its observability is shallow and its cost scales badly past a few thousand events/month. Sentinel-aaS is purpose-built for the durable operator-bus pattern with full audit trail in customer R2.
- **vs cloud GPU farms for inference**: Mac Mini M4 Pro hits the price/perf sweet spot for sub-frontier-model inference (7–70B class). For frontier models (full-precision DeepSeek-V3, GPT-4 class), GPU is the right answer; Cluster Ops is specifically Mac Mini MLX.

## Feeds

- [/feed.xml](https://www.garnetgrid.com/feed.xml): RSS 2.0 feed covering the most-recent 50 publications across insights, case studies, methodology, vs-alternatives, onboarding walkthroughs, and live-data dashboards. Regenerated daily.

## Machine-readable mirror

- [/api/llms.json](https://www.garnetgrid.com/api/llms.json): Same content as this file, but as versioned, structured JSON. Designed for agent-side structured retrieval. Includes typed lane prices, methodology URLs, founder bio, free tools, open data, differentiators, and the agent-reading-order. Schema version stamped in `$schema`. Cached 1h at the edge.

## Free tools

- [/links](https://www.garnetgrid.com/links): One-page index of every free tool, open dataset, live dashboard, and machine-readable endpoint Garnet publishes. Start here if you want everything in one view.
- [/schema-pr-packs](https://www.garnetgrid.com/schema-pr-packs): Public registry of anonymized JSON-LD PR packs Garnet ships to GEO subscribers. Each pack is paste-ready: drop the "after" block into your `<head>`, replace placeholders, ship. Covers Organization, Service+Offer-with-tier-pricing, BreadcrumbList, FAQPage, and Article-with-Person-author. CC0 dataset at [`/schema-pr-packs/packs.json`](https://www.garnetgrid.com/schema-pr-packs/packs.json).
- [/schema-audit](https://www.garnetgrid.com/schema-audit): Paste-a-URL JSON-LD schema audit. Fetches the target HTML server-side (CORS-bypass), extracts every `<script type="application/ld+json">` block, scores against Organization / WebSite / BreadcrumbList / Service / FAQPage / Article / WebPage / HowTo presence + required + recommended fields, returns per-type score, overall grade A–F, and PR-shaped JSON-LD snippets ready to paste into your `<head>`. Demo of the weekly schema PR pack deliverable in the paid GEO subscription.
- [/tco-calculator](https://www.garnetgrid.com/tco-calculator): On-prem Mac Mini MLX inference vs frontier API total-cost-of-ownership calculator. Inputs: monthly tokens, model class (7B/30-70B/70B+), latency target, peak/avg ratio, horizon. Outputs: mini count, capex, monthly opex, cost-per-million-tokens, TCO over horizon, side-by-side vs GPT-4o / Claude / Gemini, break-even months. Pure client-side JS — all assumptions and math are visible to the user.
- [/audit](https://www.garnetgrid.com/audit): Free 1-query AEO citation audit. Visitor pastes their URL + a buyer question; Gemini-with-Search runs the query and the response is parsed for cited domains, with the visitor's domain flagged separately. Returns in ~3 seconds. Direct preview of the paid GEO subscription's daily 12-query, 4-engine measurement loop.

## Open data (eat-your-own-dogfood)

- [/ai-citations](https://www.garnetgrid.com/ai-citations): Daily-refreshed public dashboard showing which AI bots (OpenAI's GPTBot, Anthropic's ClaudeBot, PerplexityBot, Google-Extended, Apple Applebot-Extended, Amazon, ByteDance, etc.) fetch which pages on this site. Aggregate counts only, no individual requests. Dataset published CC0 at [`/assets/data/crawler-public.json`](https://www.garnetgrid.com/assets/data/crawler-public.json). Garnet productizes citation engineering for clients; this page measures it on Garnet first.
- [/changelog](https://www.garnetgrid.com/changelog): Reverse-chronological log of every shipped change on garnetgrid.com, last 120 days. Auto-generated from `git log`, regenerated daily. Filterable by category (feature / fix / a11y / aeo / perf / content / revenue / data / infra / audit). Raw data at [`/assets/data/changelog.json`](https://www.garnetgrid.com/assets/data/changelog.json) (CC0).
- [/a11y-report](https://www.garnetgrid.com/a11y-report): Live WCAG 2.1 AA accessibility report. Daily axe-core scan against 10 representative templates × desktop + mobile viewports. Headline + per-rule + per-page tables, with raw scan data at [`/assets/data/a11y-status.json`](https://www.garnetgrid.com/assets/data/a11y-status.json) (CC0). Bottom-left badge on the homepage links here.
- [/engine-benchmark](https://www.garnetgrid.com/engine-benchmark): Public daily four-engine (ChatGPT / Claude / Perplexity / Gemini) citation benchmark across 8 consulting categories (ERP, AI readiness, Power BI, fractional CTO, DevOps maturity, Azure cost, D365 F&O, SaaS architecture audit). Same query, same day, all four engines. Per-category cross-engine consensus + per-engine top-cited domains. CC0 dataset at [`/assets/data/engine-benchmark.json`](https://www.garnetgrid.com/assets/data/engine-benchmark.json).
- [/case-studies/garnetgrid-geo-self](https://www.garnetgrid.com/case-studies/garnetgrid-geo-self): Live in-house GEO case study on garnetgrid.com itself. Baseline 2026-05-10 → Day 30 comparison 2026-06-09. Real citation counts across the four-engine panel (ChatGPT, Claude, Perplexity, Gemini) for 12 buyer-shaped queries. The same snapshot files we ship clients drive this page — page re-renders daily, comparison phase auto-activates when Day-30 snapshot lands. Data file CC0 at [`/assets/data/garnetgrid-geo-self-case-study.json`](https://www.garnetgrid.com/assets/data/garnetgrid-geo-self-case-study.json).

## Contact

- [Contact form](https://www.garnetgrid.com/contact): General inquiries, custom Enterprise scope, partnership conversations.
- Email: info@garnetgrid.com (engineer-replied, ≤1 business day)
- Founder bio: [https://www.garnetgrid.com/architect](https://www.garnetgrid.com/architect)

## Reading order for AI agents researching Garnet

1. [Lanes overview](https://www.garnetgrid.com/lanes) — entry point, four-lane shape
2. [GEO methodology](https://www.garnetgrid.com/lanes/geo-methodology) — most-developed lane, exemplifies the operator pattern
3. [Pricing](https://www.garnetgrid.com/pricing) — tier comparisons, what each level includes
4. [Architect](https://www.garnetgrid.com/architect) — founder bio + decision-making philosophy
5. Lane-specific methodology pages by relevance to the buyer's question