Shray, Solvrix AI
Shray

Me and my team hold every build to one standard: it has to outperform what the client expected when they asked for it.

Built by Us

Hey, glad to see you here. Please take a look at what we've shipped — real builds, real architecture, below.

Category 01

n8n // Workflow Orchestration

Agentic workflows built on n8n — the orchestration layer behind lead generation, ad intelligence, and deal sourcing pipelines we run for clients.

n8n · 01
Lead Discovery Agent
Fans a single search request out across ten independent lead sources in parallel — Google Search, Google Maps, LinkedIn, Apollo, Clutch, Europages, Yellowpages, Yell, 411, and Yelp. Each branch retrieves and deduplicates its own results before two merge stages collapse everything into one clean prospect list, with a final pass to catch cross-source duplicates. Runs as a sub-workflow, called on demand by the parent pipeline.
n8n Workflow
◆ Live n8n Canvas — Production Workflow
Lead Discovery Agent n8n workflow canvas showing parallel branches across ten lead sources merging into a deduplicated prospect list
Fan-Out
10 Parallel Sources
Google · Maps · LinkedIn · Apollo
Retrieval
Per-Source API Calls
HTTP Requests · SERP-Style
Cleansing
Dual-Stage Deduplication
Per-Branch + Global Pass
Merge
Two-Stage Consolidation
n8n Merge Nodes
Trigger
Sub-Workflow Call
Invoked by Parent Pipeline
n8n · 02
Multi-Channel Follow-Up Engine
Two coordinated workflows sharing one Google Sheet as the source of truth. The sequencer triggers on any sheet update, computes follow-up timing, and waits before checking each contact's email and phone status — branching to Gmail, SMS, or WhatsApp accordingly, across both a first follow-up and a final follow-up pass. A second workflow listens for inbound status updates via webhook and writes the result — Confirm, Reschedule, Stop, or Default — straight back to the sheet, keeping it as the single live record of where every contact stands.
n8n Workflow
◆ Live n8n Canvas — Production Workflow
Multi-channel follow-up engine n8n workflow showing sequencer with email, SMS, and WhatsApp branches alongside a webhook-driven status handler writing back to Google Sheets
Source of Truth
Google Sheets
Trigger · Read · Write
Sequencing
First + Final Follow-Up
Computed Wait Timing
Channels
Email · SMS · WhatsApp
Gmail · SMS API · WhatsApp API
Status Handling
Confirm / Reschedule / Stop
Webhook → Switch → Sheet Update
n8n · 03
AI Lead Engagement Agent
An n8n AI Agent — wired with memory and tool access, not just a bare LLM call — handles both voice and chat intake behind one webhook. It pulls live company context, routes each conversation to a voice or chat track, decides whether the interaction needs an AI response at all, and runs the ones that do through a shared agent with OpenAI as its model backend. On the voice side, it decides whether to escalate cross-channel and fires an SMS via Twilio when it does. On the chat side, it checks for usable lead data and writes qualified leads straight to Google Sheets — with a fallback path on both sides when there's nothing actionable.
n8n Workflow
◆ Live n8n Canvas — Production Workflow
AI Lead Engagement Agent n8n workflow showing a shared AI Agent node with memory and tools handling voice and chat routing, cross-channel SMS escalation, and lead capture to Google Sheets
Intake
Voice + Chat Routing
Webhook · Company Context Fetch
Agent Core
Memory + Tools
n8n AI Agent · OpenAI Model
Escalation
Cross-Channel SMS
Twilio
Lead Capture
Qualified Lead Write
Google Sheets
n8n · 04
Meta Ads Creative Intelligence Engine
Pulls live ad creatives from Meta via Apify, polling the actor run until results are ready, then splits the batch into video and image tracks — each deduplicated before processing. Every asset loops through individually: downloaded, run through AI analysis, and logged as a structured row in Google Sheets. Both tracks respond back to their originating webhook the moment their loop finishes, so the caller knows the batch is done without needing to poll separately.
n8n Workflow
◆ Live n8n Canvas — Production Workflow
Meta Ads Creative Intelligence Engine n8n workflow showing Apify ad retrieval, video and image deduplication loops, AI analysis steps, and Google Sheets logging
Retrieval
Apify Actor Run
Request → Poll → GetResult
Cleansing
Video + Image Dedup
Per-Track Loops
Intelligence
Per-Asset AI Analysis
Video + Image Analysis
Output
Structured Sheet Log
Google Sheets · Dual Webhook Reply
Category 02

GoHighLevel // Client-Facing Automation

CRM workflows and pipeline automation built natively inside GoHighLevel — the system clients log into directly.

GHL · 01
Customer Onboarding & Post-Purchase Automation
The moment a purchase tag lands on a contact's record, this workflow picks it up and takes over. It watches for any contact creation or update, then checks which product or bundle tag was just applied — AI playbook, AI agent bundle, Manychat, and others each have their own dedicated branch — and immediately routes the new customer into the correct onboarding sequence for what they actually bought. No manual handoff between checkout and onboarding: the tag is the trigger, and the right welcome sequence starts automatically every time.
GoHighLevel Workflow
◆ Live GoHighLevel Builder — Production Workflow
GoHighLevel customer onboarding workflow showing a single condition node branching by purchase tag into multiple parallel onboarding paths
Trigger
Purchase Tag Applied
Contact Created or Changed
Decision
Product-Specific Routing
Tag-Match Fan-Out
Segmentation
Per-Product Onboarding
Bundle · Plan Tags
Output
Automatic Welcome Sequence
No Manual Handoff
GHL · 02
Lead Nurturing & Conversion Workflow
Once a contact lands in a branch, the work doesn't stop at one email — each path runs its own multi-touch nurture sequence, spacing messages out with built-in wait times so contacts get a steady drip instead of a single blast. Some branches repeat the same offer to reinforce it; others rotate between a feature announcement, a Manychat promotion, and a strategy call upsell to keep the conversation moving toward a decision. Every step stays tagged to a specific audience, so the right contacts stay in the right conversation all the way to conversion — without anyone manually tracking who's seen what.
GoHighLevel Workflow
◆ Live GoHighLevel Builder — Production Workflow
GoHighLevel lead nurturing workflow showing multiple tag-based branches each running a multi-step drip sequence of waits and emails
Cadence
Multi-Touch Drip Per Branch
Wait → Send → Wait → Send
Messaging Logic
Reinforce or Rotate Offers
Feature · Promo · Upsell
Targeting
Audience Stays Tagged
No Manual Tracking
Outcome
Relevance at Scale
Toward Conversion
GHL · 03
Seller Lead Intake & Routing
Triggers off a seller lead survey submission and immediately tags, sources, and places the lead into the right CRM pipeline before assigning it to a team member. After a wait period, a quality check reads the lead's stated timeline — leads wanting to sell within 30 days get tagged hot, trigger an internal notification, and generate a call task for immediate follow-up. Everyone else gets tagged warm, added to a nurture workflow, and queued for a scheduled follow-up task instead. One trigger, two distinct treatment paths based on actual lead intent.
GoHighLevel Workflow
◆ Live GoHighLevel Builder — Production Workflow
GoHighLevel Seller Lead Intake and Routing workflow showing survey trigger, tagging, pipeline placement, and a hot/warm lead quality branch
Scroll to see full workflow
Trigger
Survey Submission
Seller Lead Form
CRM Setup
Tag · Source · Pipeline
Auto-Assign to User
Scoring
Timeline-Based Quality Check
Hot vs. Warm Branch
Action
Notify + Task + Nurture
Email · SMS · Call Task
Category 03

Zapier & Make // Integration & Decision Logic

Fast, lightweight integrations connecting booking systems, CRMs, and spreadsheets — built for speed without sacrificing data integrity.

Zapier · 01
Appointment-to-CRM Sync & Data Completion
Every new Acuity appointment kicks off a sync into both Agile CRM and Klaviyo, then pauses before checking a Google Sheet for that contact's existing record. The system makes two separate decisions in sequence — first whether an email is already on file, then whether a phone number is — and routes accordingly each time: update the existing row if the data's there, create a new one if it isn't. The result is a spreadsheet that never ends up with duplicate or incomplete contact records, no matter how a lead originally came in.
Zapier Workflow
◆ Live Zapier Builder — Production Workflow
Zapier workflow syncing Acuity appointments to Agile CRM and Klaviyo, then checking Google Sheets for existing email and phone data before updating or creating rows
Trigger
New Appointment
Acuity Scheduling
Sync
Dual CRM Update
Agile CRM · Klaviyo
Decision 1
Email On File?
Update vs. Lookup Further
Decision 2
Phone On File?
Update vs. Create Row
Zapier · 02
Acuity-to-HubSpot Orientation Sync
Every time someone books an orientation through Acuity, this Zap checks HubSpot right away to see if that person already exists as a contact. Already in the system? Their record gets updated, not duplicated. Brand new? HubSpot creates the contact on the spot. Either way, by the time the orientation starts, the lead is already sitting in the CRM, accurately recorded, with zero manual data entry.
Zapier Workflow
◆ Live Zapier Builder — Production Workflow
Zapier workflow named Acuity Booked to HubSpot Orientations, finding a HubSpot contact then branching to update or create based on whether the contact already exists
Trigger
New Orientation Booking
Acuity Scheduling
Lookup
Find Existing Contact
HubSpot
Decision
Contact Exists?
Single Branch Point
Outcome
Zero Manual Data Entry
Update or Create Contact
Zapier · 03
Approved Document Auto-Delivery
A coach finishes a deliverable, marks the row "approved" in Google Sheets, and that's the last manual step. The Zap picks it up immediately: a filter confirms the approval and status are actually valid, a code step extracts the document ID, Google Drive exports the file, and Gmail sends it straight to the client — no coach has to remember to attach or send anything. The final step writes the status back to the sheet as sent, so the same row never gets processed twice. One checkbox in, one delivered document out.
Zapier Workflow
◆ Live Zapier Run History — Production Workflow
Zapier run history showing an approval-gated document delivery workflow: Google Sheets approval, filter check, document ID extraction, Google Drive export, Gmail send, and status update back to the sheet
Trigger
Approval Marked in Sheet
Google Sheets
Gate
Approval + Status Filter
Prevents Invalid Sends
Delivery
Export + Email to Client
Google Drive · Gmail
Safeguard
Status Write-Back
No Duplicate Sends
Make
Bubble-to-Sheets Job Sync Engine
On a repeating schedule, this system pulls live job data straight from a Bubble.io app, runs it through a custom filtering step, and then makes a series of decisions about where each job record actually belongs. Active postings get inserted into Care.com job sheets; cancelled ones get logged separately so they never get confused with live listings. Some branches go a step further, generating a fresh templated sheet copy and looping through multiple rows automatically. Every path — no matter which route a job takes — eventually funnels back through a final routing point, so nothing falls through the cracks and every job ends up exactly where it should, without anyone touching a spreadsheet by hand.
Make Workflow
◆ Live Make Builder — Production Workflow
Make.com scenario pulling job data from Bubble.io, filtering it through custom logic, and routing active versus cancelled jobs into Google Sheets through multiple nested routers and iterators
Source
Bubble.io API Pull
Scheduled Repeater
Logic
Custom Filter Step
Code Module
Routing
Nested Multi-Level Routers
Active vs. Cancelled Jobs
Output
Sheet Sync + Templating
Google Sheets · Iterator
Category 04

Custom Builds // RAG & Bespoke Systems

Builds that don't fit neatly into a single no-code platform: custom dashboards, retrieval-augmented knowledge systems, and bespoke automation layered on top of a client's own data.

Custom · 01
WhatsApp Sales Agent & Property RAG
A WhatsApp sales agent built for a real estate client, backed by a custom knowledge base pulled from their live property listings on Drive. Customers message in with questions about a listing or which property fits what they're looking for, and the agent answers and recommends straight from that knowledge base inside the chat, no manual lookup needed on the client's end. Every conversation is scored for lead quality as it comes in, with property matching, media delivery, and booking intent handled in the same thread. Anything the agent can't fully resolve gets routed to a human takeover queue instead of going unanswered, and the dashboard tracks active conversations, site visits booked, and follow-up conversions in one place.
Custom Build
◆ Live Automation Dashboard — Production Workflow
WhatsApp Sales Agent and Property RAG automation dashboard
Intake
WhatsApp Webhook Inbound
Dedicated WhatsApp Number
Knowledge
Property Retrieval
Drive RAG · Live Listings
Qualification
Lead Quality Scoring
Hot Lead Detection
Escalation
Human Takeover Queue
Operator Dashboard
Custom · 02
Script-to-Reel AI Avatar Pipeline
A script-to-reel content pipeline built for a real estate client, covering everything from content angle to a ready-to-post Instagram reel. The team picks a content type and location, like Commercial Investment listings in Panchkula, and the system drafts a Hinglish, voice-safe script tuned to the brand's tone from a custom brief. Once the draft is approved, it converts to voice, then hands off to HeyGen to generate the talking-head AI avatar video. The final reel comes out ready for manual Instagram posting, with the whole script-to-voice-to-approval flow tracked through one dashboard.
Custom Build
◆ Live Content Studio — Production Workflow
Script-to-reel AI avatar content pipeline dashboard
Input
Content Angle Selection
Content Type · Location
Draft
AI Script Generation
Hinglish · Custom Brief
Voice
Approved Draft to Voice
Voice-Ready Conversion
Render
AI Avatar Video
HeyGen Handoff
Category 05

Architectural Designs // Conceptual Intelligence Frameworks

Four core intelligence frameworks designed for CPA firm operations. Each blueprint shows the complete data flow — from trigger through collection, reasoning, memory, confidence scoring, and output.

The Stack // Technology We Deploy

Every architecture below is built on the same four-layer technology stack — chosen specifically for data sovereignty, accuracy, and enterprise reliability in accounting environments.

n8n Enterprise
Orchestration Layer · Workflow Engine
Claude API
Reasoning Layer · Agentic Logic
Pinecone
Memory Layer · Vector Intelligence
RAG Architecture
Knowledge Retrieval · Context Engine
Confidence Scoring
Accuracy Gate · Human Review Routing
Private Cloud
Client Environment · Zero Data Leakage
Blueprint 01
Document Intelligence Pipeline
Automates the ingestion, extraction, and processing of client documents — tax returns, contracts, financial statements, client intake forms. The agent reads, extracts structured data, cross-references against firm knowledge, scores confidence, and either acts autonomously or routes to human review.
Architecture Blueprint
◆ Conceptual Architecture Diagram — Implementation detail defined in Forensic Report
TRIGGER Doc Received EMAIL / UPLOAD LAYER 1 Collection PDF · XLSX · EMAIL n8n Ingestion LAYER 2 Extraction OCR · PARSE · MAP Claude API LAYER 3 Confidence SCORE · THRESHOLD 0–100% Accuracy LAYER 4 RAG Query VECTOR SEARCH Pinecone OUTPUT Action / Flag AUTO OR REVIEW MEMORY Client SOPs Vector DB KNOWLEDGE Tax Codes · Rules Pinecone Index
Collection
Document Ingestion
n8n · Email · Upload
Extraction
OCR + Structured Parse
Claude API · Pattern Match
Intelligence
Confidence Scoring
0–100% Accuracy Gate
Memory
RAG Knowledge Query
Pinecone · Tax Codes · SOPs
Output
Auto-Action or Flag
CRM Update · Alert · Log
Blueprint 02
Reconciliation Intelligence Agent
Continuously monitors financial data across connected accounts, detects anomalies, cross-references against historical patterns and compliance rules, applies a confidence gate, and either auto-resolves or escalates with full context. Every action is logged to an immutable audit trail.
Architecture Blueprint
◆ Conceptual Architecture Diagram — Implementation detail defined in Forensic Report
SYNC Data Pull QBO · XERO · SHEETS LAYER 1 Anomaly Detection n8n Logic Engine LAYER 2 RAG Cross- Reference Pinecone · Claude GATE Confidence Threshold >95% AUTO · else FLAG LAYER 3 Auto-Resolve Reconcile + Log OUTPUT Audit Log Full Trail · Sheets LOW CONF Human Review Email · Slack Alert FLAGGED Senior Review Priority Alert MEMORY LAYER — CLIENT RULES · HISTORICAL PATTERNS · FIRM SOPs
Sync
Live Data Pull
QBO · Xero · Sheets
Detection
Anomaly Engine
n8n Logic · Thresholds
RAG
Cross-Reference
Pinecone · Claude · History
Gate
Confidence Threshold
>95% Auto · else Human
Audit
Immutable Log
Full Trail · Sheets · CRM
Blueprint 03
Client Intelligence Layer — 5-Memory Architecture
The most sophisticated layer in the stack. A five-tier memory architecture that gives every agent full context about each client — their history, preferences, compliance profile, and all past interactions. The reasoning agent queries all five memory layers before taking any action, scores its confidence, and routes accordingly.
Architecture Blueprint
◆ Conceptual Architecture Diagram — Implementation detail defined in Forensic Report
5-LAYER MEMORY ARCHITECTURE TRIGGER CRM Event NEW CLIENT · TASK LAYER 1 Context Retrieval Client History · Profile n8n + CRM LAYER 2 RAG Memory Query SOPs · Preferences Pinecone Vector LAYER 3 Reasoning Agent CoT · Business Rules Claude API LAYER 4 Confidence Scoring Score · Route · Gate 0–100% Threshold LAYER 5 Draft + Action Email · Update · Log n8n Execute GATE Human Review M1 Client Profile CRM Data M2 Interaction Log All History M3 Firm SOPs Vector Index M4 Tax Codes Compliance Rules M5 Decision Log Agent History PINECONE Private Vector Database Sovereign · Encrypted · Yours
Memory 1
Client Profile
CRM · Structured Data
Memory 2
Interaction Log
Full Comms History
Memory 3
Firm SOPs
Pinecone Vector Index
Memory 4
Tax Codes + Rules
Compliance Layer
Memory 5
Decision History
Agent Learning Log
Blueprint 04
Deadline & Compliance Monitoring Agent
Runs a daily scan across all active clients, loads full client context from the memory layer, checks compliance requirements against current tax codes, scores urgency by days remaining, and generates priority-ranked alerts pre-loaded with everything the accountant needs to act. Urgent deadlines escalate directly to senior partners.
Architecture Blueprint
◆ Conceptual Architecture Diagram — Implementation detail defined in Forensic Report
TRIGGER Scheduled DAILY SCAN · CRON LAYER 1 Deadline Scan All Clients · All Dates n8n · CRM Pull LAYER 2 RAG Context Load Client Profile · History Pinecone · Memory LAYER 3 Compliance Check Tax Codes · Rules · IRS Claude API · RAG GATE Priority Scoring URGENT · HIGH · LOW Days Remaining LAYER 4 Alert Generation Email · Slack · SMS Context Pre-Loaded LOG Audit Trail URGENT · <7 DAYS Senior Partner Alert Immediate Escalation MEMORY LAYER Filing Calendar All Client Deadlines Compliance Rules IRS · State · Federal Alert History Sent · Acknowledged Client Risk Profile Late Filing History
Trigger
Daily Cron Scan
n8n Scheduler
RAG
Client Context Load
Pinecone · CRM · Memory
Compliance
Rules Verification
Claude API · IRS · State
Gate
Priority Scoring
Urgent · High · Low
Output
Contextual Alert
Email · Slack · Escalation