Real Client Work
Client Project
Chris Buys Houses - Full Operations Backend
Chris needed his lead management tool (DataSift) to talk to his sales CRM (GoHighLevel) in both directions. When a lead's status changed in one system, the other had to update automatically. Tags, pipeline stages, sold status, attempt tracking - all of it.
- Two-way sync between REISift/DataSift and GoHighLevel
- Automated lead tagging across two priority tiers with overlap dedup
- Pipeline stage sync (attempts, dispositions, sold status) flowing both directions
- Token manager for API authentication refresh
- Internal Slack notifications on key pipeline events
- Bulk status update handler for batch operations
This wasn't one workflow. It was 15 interconnected pieces that had to stay in sync without breaking each other.
Demo Builds
Fully functional workflows I built as demonstrations. Real APIs, real data routing, real AI processing. Not mockups.
Demo Build
Client Onboarding Automation
A new client fills out an intake form. From there, everything happens automatically:
- Creates a project folder structure in Google Drive
- Spins up a Trello card with a pre-built checklist
- Sends a branded HTML welcome email to the client
- Logs the onboarding event to Google Sheets
- Posts a summary to the team's Slack channel
- Error handling on every external node so nothing fails silently
One form submission. Five systems updated. Zero manual work.
Demo Build
Shopify Custom Order Intake
Built for a custom product shop (demo brand: "Timber & Thread Co."). When a Shopify order comes in with customization details buried in the order notes field:
- DeepSeek AI extracts the specs (size, material, color, engraving text, whatever the customer typed)
- Creates a structured production task in Airtable with all specs broken out
- Sends a branded HTML confirmation email to the buyer with their customization details listed back to them
The hard part isn't the routing. It's getting AI to reliably parse freeform customer text into structured fields that a production team can actually use.
Demo Build
Revision Request Tracker
Monitors a Gmail inbox for emails that look like revision requests. When one comes in:
- AI parses the email to extract what the client wants changed, which project it's for, and how urgent it is
- Logs it to Google Sheets with all fields structured
- If the AI flags it as urgent, fires a Telegram ping to the project owner immediately
- Sends an acknowledgment email back to the client so they know it was received
The AI classification is the key piece - it has to distinguish "hey just checking in" from "this needs to be fixed before our launch tomorrow."
Demo Build
Inbox to CRM Pipeline
Not every email deserves CRM attention. This workflow:
- Watches Gmail for new messages
- Filters out marketing, newsletters, and noise automatically
- Runs the remaining emails through DeepSeek to extract sender, intent, urgency, and relevant details
- Creates a structured entry in Notion with all that context
- Posts a clean summary to Slack so the owner can scan it without opening email
The filtering is what makes this useful. Without it, you're just dumping every email into a CRM and making more noise, not less.
Demo Build
Lead Triage System
When a lead comes in through any web form:
- Normalizes the form data regardless of which form platform sent it
- DeepSeek scores the lead on quality (budget signals, urgency, fit)
- Logs to Google Sheets with the score and reasoning
- Fires a Slack alert with the score so the sales person knows whether to prioritize it or let it sit
Saves the first 5-10 minutes of every inbound lead that would otherwise go to manual review.
Systems I Built for My Own Business
I run an automation agency. These are the internal systems I built to run it. They're not demo projects - they run daily in production.
Production System
Cold Outreach Engine
A Google Maps scraper feeds prospect data into a two-account Gmail outreach system with automated follow-ups. The whole pipeline runs on its own.
- Google Maps scraper pulls business data by niche and location - a free alternative to paid lead databases
- AI writes each cold email from scratch based on the prospect's business - no templates
- Two Gmail accounts with staggered send windows (cold: 8-10am, FU1: 10am-12pm, FU2: 12-2pm)
- Cross-account dedup so the same prospect never gets emailed twice
- Row-level tracking in Google Sheets (even rows = Account A, odd = Account B)
- Bad email format detection and auto-skip
- Five rotating templates per follow-up stage
Production System
Facebook Lead Scout
Passively monitors 20 Facebook groups for potential clients:
- RSS scanner watches all 20 groups on a loop
- AI classifier reads every post and decides GO, MAYBE, or SKIP based on buying signals
- Dedup sheet so no lead gets surfaced twice
- Telegram alerts on every GO with context and the direct link to the post
The classifier prompt alone went through 10+ iterations to cut false positives from ~90% down to a usable rate. Runs in the background while I work on other things.
