I design AI-powered dashboards, workflow automations, CRM systems, reporting platforms, internal tools, and operational software for businesses that need greater visibility, efficiency, and control. Everything on this site is real work for real businesses — every project carries an honest status label.
These are the only businesses on this site, because they are the only businesses I currently work for. No hypothetical brands, no fabricated testimonials, no borrowed logos.
Flagship product. Multi-business marketing, analytics, CRM, messaging, and operations platform. Next.js + Supabase, deployed on Vercel, running in production today.
app.cloutsyndicate.io → full case study ↓
Automotive retail marketing, lead intelligence, and performance operations. Google Ads restructuring, Meta lead generation, real-time lead capture and routing.
Diesel performance marketing operations and an AI-assisted lead-response system: 24-entry approved knowledge base, follow-up sequences, SLA escalation, technical-accuracy workflow.
Automotive service marketing and growth operations. Campaign relaunch, service-offer structure, brake/AC/inspection promotion, reporting integration.
Local growth, membership marketing, and retention systems. Meta campaign history integrated; membership and retention programs in planning.
CREDIBILITY STANDARD
Every claim on this site is interview-safe: accurate, defensible, and labeled with its real stage. Customer PII never appears — dashboards shown use sanitized or demonstration data.
A centralized multi-business marketing, analytics, CRM, automation, and operations platform — designed, built, and operated for the five businesses above. Live in Production
STACK
Next.js 16 · React 19 · TypeScript · Supabase Postgres · Vercel · Tailwind
ROLE
Product design, architecture, build, integration, deployment, operations — AI-assisted development, human-owned decisions
SCALE
5 businesses · 22 API routes · 30+ tables · 14 migrations · 4 scheduled jobs · 4 webhook integrations
DATA INTEGRITY
Ad spend reconciled to the dollar against Google Ads for all three connected accounts (July 2026 audit)
Managing marketing for multiple businesses means living across Google Ads, Meta Ads, lead forms, DMs, SMS, email, spreadsheets, and reporting tools. The result: repetitive logins, manual reporting, disconnected data, slow decisions, and — the finding that started this build — leads that got slow responses or none at all, with no system that even noticed.
The Command Center is one operating environment where an administrator oversees every business, each business stays cryptographically separated at the database layer, clients see only their own data, and AI assists with drafting, summarizing, and alerting — always behind human approval.
Before any code, a full audit of the existing lead flow found the real root causes — none of which were "we need AI":
Each was fixed as a deterministic correction — before any AI feature was considered. Lesson: most "AI problems" are actually plumbing problems.
One platform, many businesses, with separation enforced in the database — not in the UI.
Status nuance, stated plainly: the AI response layer is implemented and safety-gated but not yet activated (pending API key + owner's knowledge-base approval). Content calendar and alert-rules screens currently render clearly-labeled sample data pending their integrations.
AI is used where judgment-at-scale helps, never where determinism is required:
Before demoing this platform to anyone, I audited my own product against the source platforms:
MVP shipped first: secure auth · business separation · role-based access · core dashboard metrics · Google Ads + Meta reporting · client-specific views · lead capture and delivery · basic alerts · downloadable weekly reports. Everything else was refused until these were verified.
Live today: everything in the MVP plus two-way messaging plumbing, SLA escalation on a 5-minute clock, knowledge-base approvals, attribution confidence, weekly executive PDFs.
Roadmap (labeled honestly): AI response activation (pending key + approvals) · SMS channel (carrier registration re-file in progress) · outbound Meta DMs (verification confirmed, portfolio wiring in progress) · GA4 + Search Console integration (planned) · call tracking (schema built, provider not wired) · content calendar integration (planned) · cross-client benchmarking (concept).
Each study follows the same discipline: real problem → current-state workflow → requirements → solution → MVP → AI role → human oversight → measurement → honest status. Metrics shown are previously published campaign figures or qualitative observations — nothing invented.
AUTOMOTIVE RETAIL MARKETING · LEAD INTELLIGENCE · PERFORMANCE OPERATIONS
Problem. A used-car dealership generating leads across Google Ads, Meta lead forms, and Instagram/Facebook DMs — with fragmented visibility, manual reporting, and lead follow-up living in one person's inbox.
Work performed (verified): Google Ads account restructuring and campaign QA · Meta lead-form campaigns with real-time webhook capture into the CRM · lead delivery with multi-recipient routing and delivery audit · executive weekly reporting · attribution capture (campaign/ad per lead) · cross-channel performance analysis.
Published results: 124 client-intake leads in a 30-day campaign window Previously published · cloutsyndicate.io
AI opportunities (proposed, labeled): lead-priority summaries · bilingual response drafting · inventory promotion recommendations · automated campaign-health reports. Concepts
DIESEL PERFORMANCE MARKETING OPERATIONS · AI-ASSISTED LEAD RESPONSE
Problem. A technical diesel shop where some leads got delayed responses and others got none — and where wrong technical claims in marketing carry real liability.
Work performed (verified): full lead-flow audit · AI knowledge base (24 owner-approved entries: services, Banks Power/PPE installer status, pricing policy, compliance refusals) · 5 follow-up sequences · 60-minute SLA with two-tier escalation, live on a 5-minute clock · website rebuild and migration to Cloudflare Pages · Google Ads management · weekly reporting.
Published results: 69 client-intake leads in a 30-day campaign window Previously published · cloutsyndicate.io
The technical-accuracy workflow (this is the part that matters for AI content): identify exact part/service → verify vehicle compatibility → research real specs → approved brand assets only → validate claims → human review → owner approval → publish → measure. The AI refuses emissions-delete work outright — a compliance rule enforced in code, decided by the owner, not left to a model.
Honest status: responder implemented and safety-gated; activation awaits final credentials and carrier registration (SMS re-file in progress, root cause diagnosed and corrected). In Development
AUTOMOTIVE SERVICE MARKETING · CUSTOMER COMMUNICATION · GROWTH OPERATIONS
Problem. A local repair shop competing on service offers — brakes, AC, diagnostics, inspections — needing consistent promotion, conversion tracking, and reporting without an internal marketing team.
Work performed (verified): campaign restructure and July relaunch (search campaigns for auto repair & diagnostics, brand, maintenance offers, warranty repair) · offer architecture · reporting integration into the Command Center · lead capture via Meta forms.
Before → after workflow: separate platforms, manual reporting, inconsistent visibility → centralized dashboard, standardized weekly reports, SLA-tracked lead handling. The after-state is live for reporting and leads; AI-assisted service recommendations and review-response drafting are labeled concepts. Concepts
LOCAL GROWTH · MEMBERSHIP MARKETING · RETENTION SYSTEMS
Problem. A multi-service wash business (washes, tint, wrap, PPF, ceramic, detail) whose growth depends on memberships and repeat visits, not one-time transactions.
Work performed (verified): Meta advertising with campaign history integrated into the Command Center's metrics warehouse · business onboarded into the multi-tenant platform with its own pipeline and branding.
Planned (labeled): membership campaign strategy · seasonal calendars · retention alerts · weather-triggered promotion concepts · reporting integration (currently the one business intentionally excluded from the weekly report cycle — that gap is documented, not hidden). Planned
No decorative complexity. Each diagram documents a decision that exists in the running system.
Why it matters: business separation survives application bugs because it is enforced a layer below the application. The internal gate originally failed open when unconfigured — found in audit, now fails closed.
Daily API syncs (Google Ads campaign-level, Meta account-level) + real-time webhooks → single metrics warehouse keyed by business/date/source/level → live dashboard queries with zero caching. Dedup enforced by unique indexes; Meta window uses delete-then-insert.
5-minute heartbeat → find conversations breaching 60-minute response target → tier-1 email to three recipients → tier-2 escalation at 120 minutes → 1-day stale SMS. Only a human reply clears the clock — AI acknowledgments don't count as answered.
Platform-reported IDs (campaign/ad) captured at webhook time; website UTM/click-ID capture; every lead stamped Verified / Probable / Unknown / Manual. Unknown is displayed as unknown — never guessed.
Metrics warehouse → weekly comparison engine (7d vs prior 7d) → per-business executive PDF with agency + client variants. One shared metric-definition layer is the documented next step (report and dashboard formulas are being unified).
AI runs, handoffs, approvals, assignments, and outcomes all write audit rows with actor + timestamp. The knowledge base is versioned — every approved answer knows who approved it and when.
GitHub → Vercel (build-per-commit, promote-on-green) → Cloudflare DNS. Static client sites on Cloudflare Pages. Rollback = promote any previous green build. Secrets live in platform env vars, never in code.
The experimentation layer behind the products. Everything here connects to the five real businesses above — no confidential client data, no invented results.
Structured prompts with fixed sections — role, business context, approved claims, refusal rules, output schema — used for campaign strategy, creative briefing, QA, reporting summaries, and client-safe communication. The production responder's prompts draw only from an owner-approved knowledge base, with brand claims allow-listed (e.g., authorized-installer status verified with the owner before the AI may say it).
Report data → executive summary. Lead information → follow-up draft. Performance anomaly → alert. Client request → structured task. Research → verified campaign claims → creative output. Each workflow documents its inputs, its human checkpoint, and what the AI is not allowed to decide.
AI output ships only after validation: brand review, technical review (diesel parts have compatibility facts; wrong answers cost real money), legal-sensitivity screens (emissions compliance is a hardcoded refusal), and owner approval for anything customer-facing. Final accountability is always a named human.
Reporting Agent, Campaign QA Agent, Client Summary Agent, Brand Compliance Agent, Workflow Coordinator. Designed as orchestrations of the existing approved-knowledge + confidence-gate + audit-log primitives — which is why they're credible concepts rather than slideware. None are in production; all are labeled.
Run against the real systems, with honest outcomes: structured outputs vs free text for lead extraction (structured won — retries on schema mismatch beat prompt pleading) · hallucination reduction by restricting claims to an approved knowledge base with an explicit "hand off if unsure" path · confidence thresholds for when a draft is safe to send vs escalate · self-audit discipline: the July 2026 data-accuracy audit treated my own product as the system under test and found real defects. Numbers and write-ups available on request; nothing is quantified here that wasn't measured.
Twelve phases, in order, with the discipline points that actually matter. The Command Center followed every one of these — including the phases that said "don't build yet."
01
Identify the actual business need. Never assume AI is the answer — the diesel shop's core problem was leads reaching nobody, which is plumbing, not intelligence.
02
Map the current process, systems, manual steps, and failure points. The pre-build audit found four silent failure modes before a line of product code.
03
Stakeholder interviews, security and access constraints, and the questions owners must answer themselves — pricing policy, brand claims, compliance stance.
04
Time saved, faster response, fewer errors, fewer platform switches. The diesel shop's number: every lead answered inside 60 minutes, enforced by a clock, not a promise.
05
Process fix, automation, dashboard, integration, or AI — in that order of consideration. Sometimes the answer is "no new technology."
06
Smallest version that tests the riskiest assumption. Auth, separation, core metrics, and lead delivery shipped first; everything else waited.
07
Wireframes, data models, architecture diagrams, working proofs-of-concept — labeled as prototypes, never passed off as production.
08
Real stakeholders, real data. Reading the live user table caught a role misconfiguration that would have broken launch day.
09
Feedback, errors, and usability drive changes — with a typecheck-clean bar and an honest changelog.
10
Access controls, fail-closed gates, kill switches, rollback plans, monitored crons. Deploys promote on green only.
11
Reconcile against source-of-truth platforms. The July audit matched dashboard spend to Google Ads to the dollar — and documented the one ±1 tolerance.
12
Products are living systems: audits recur, definitions stay codified, and every metric keeps naming its source.
Each card says what the technology does here, where it's actually used, and my honest level with it.
The Command Center's application layer — App Router, server components, 22 API routes, edge middleware auth. Used in Projects · AI-assisted development, human-owned architecture.
System of record: 30+ tables, Row-Level Security for multi-tenancy, versioned migrations, auth. RLS is the reason client separation survives application bugs. Used in Projects.
App hosting with per-commit builds and scheduled functions (Vercel); DNS + static client sites on Pages (Cloudflare). Learned the platform limits the hard way — documented in the case study. Used in Projects.
Claude drives the response system (drafting, summarization, structured outputs) and AI-assisted development; multi-model experience across creative and analysis workflows. Used in Projects · prompt systems in production use.
Daily campaign-level sync via GAQL into the metrics warehouse; OAuth refresh-token flow; reconciled against the platform UI. Used in Projects.
Real-time lead-form and DM ingestion with HMAC signature verification, attribution capture, insights sync. Used in Projects.
Two-way SMS and email: inbound webhooks with signature checks, outbound sending, carrier compliance (A2P 10DLC registration — including diagnosing and re-filing a rejected campaign). Used in Projects.
The integration fabric: secret-bearer machine auth, HMAC payload verification, session tokens, per-zone auth topology. Working knowledge, applied daily.
Next data sources for the warehouse. Labeled honestly: designed for, not yet connected. Architecture familiarity.
Three tiers, never mixed: measured results carry a source; observed improvements are qualitative but real; expected outcomes are goals, not claims.
MEASURED
124 intake leads (RAW) and 69 (SNDP) in 30-day campaign windows; 193 total; 92% message response rate. Previously published · cloutsyndicate.io
Dashboard spend reconciled to the dollar vs Google Ads, 3 accounts, Jul 1–27 2026. July 2026 audit
OBSERVED
Eliminated platform-switching for daily review · standardized weekly reporting across businesses · centralized lead handling with delivery audit · repeatable QA and approval processes · a scalable multi-tenant architecture supporting five businesses.
EXPECTED (GOALS)
Every lead answered within 60 minutes once AI response activates · SMS + DM channels joining email · attribution coverage climbing as click-ID capture deploys · measured time savings once baseline instrumentation completes.
AI Product Manager · AI Operations · Automation & Business Systems — turning real business challenges into practical systems, workflows, dashboards, and automation.
Operations first, AI second. A year and a half — and counting — as an Account Executive / Freight Broker at ITS Logistics built the operational instincts: enterprise logistics in a fast-paced environment, time-sensitive coordination across carriers and shippers, translating customer requirements into executable plans while protecting revenue and service performance. CLOUT SYNDICATE turned those instincts into products: a multi-tenant platform, AI-assisted response systems, and reporting infrastructure for five real businesses. The combination is the point: I've lived the operational problems that AI products are supposed to solve — brokerage by day, building the systems by night.
Not claimed: a senior software engineering title. What is claimed — and demonstrated above — is product thinking, systems design, working full-stack delivery, and honest operational judgment.
Hiring managers, technology teams, and business owners welcome. The first conversation is about your workflow, not a pitch.
RENO, NEVADA · LOCALLY OWNED · CLOUT SYNDICATE LLC (A1 CREATIVES LLC)