Ayub Muhabbatzoda
AI Automation Architect | Make, n8n, Zapier & Python
New to Contra
Ayub is ready for their next project!
Most agencies handle SMS, WhatsApp, and voice calls as three separate, disconnected workflows. I'm building the system that makes them one.
An agency serving clients across multiple industries needed a unified way to respond to leads, process documents, and run follow-ups — regardless of which channel a lead came in on. Manual handling meant slow response times and inconsistent quality as they scaled to more clients.
I'm architecting an n8n automation core that connects Twilio (SMS), WhatsApp Business API, and Retell.ai (http://Retell.ai) (voice AI) into a single pipeline feeding OpenAI for instant, on-brand lead response and automated document processing — with reusable modules the agency deploys for each new client vertical without rebuilding from scratch.
Result so far: lead response time cut from hours to minutes, with a growing library of white-labeled automation modules the agency reuses across onboarding after onboarding.
This is an active, ongoing engagement — the kind of multi-channel AI automation work I specialize in, combining n8n, Twilio, WhatsApp, voice AI, and OpenAI into systems that scale with the business, not against it.
1
45
Sometimes the fix isn't a new system — it's finding the one wrong module in an existing one.
A client's Make.com (http://Make.com) scenario was supposed to combine multiple Google Drive files into a single text payload and run one Claude API call on the result. Instead, it was firing the Claude API once per file — burning API costs and returning fragmented, inconsistent output.
I traced the issue to the scenario's aggregation logic: files were being processed individually instead of being merged before the API step. I redesigned the module structure — Drive Search, Download, Array/Text Aggregator, Set Variable, then a single Claude API call — with the exact aggregator configuration and mapping formulas needed to guarantee one clean, ordered payload per run.
Result: 4x fewer API calls, consistent output, and zero need to rebuild the scenario from scratch. Delivered same day.
This is the kind of workflow debugging I do constantly across Make, n8n, and Zapier — if your automation is technically "working" but doing something expensive or wrong under the hood, I can usually find it fast.
1
43
No native way to track affiliate performance inside Fresha — so I engineered one from scratch.
A beauty & wellness booking business needed accurate attribution for affiliate-driven bookings, but Fresha (their scheduling platform) offered no built-in tracking, and their affiliates couldn't be paid fairly without real data.
I designed an end-to-end system: Dub.co (http://Dub.co) affiliate links feed into a dynamic Framer pre-lander that detects the affiliate ID and displays the correct coupon code, Fresha's coupon codes serve as the source of truth for bookings, and a Google Sheets ledger auto-calculates commission from monthly CSV exports.
Result: complete click-to-payout attribution, zero guesswork on affiliate performance, and full documentation so the client's team can onboard new affiliates without touching a line of code. Delivered fixed-price, in 5 calendar days flat.
If your booking platform, CRM, or affiliate program needs custom tracking that off-the-shelf tools can't provide, I build exactly this — with Make, n8n, Zapier, and Python.
Tools: Dub.co (http://Dub.co) · Framer · Fresha · Google Sheets · Python · Affiliate Attribution · API Integration
1
57
Zapier said it couldn't be done — so I built around it.
A recruitment agency was manually logging every call outcome into Less Annoying CRM after each JustCall conversation. Zapier's own triggers ("Call Updated" and "Call Completed") fire as two separate events and can't be linked by Call ID — so the automation everyone assumed was simple turned out to be technically impossible with default, off-the-shelf tools.
I diagnosed the exact limitation inside JustCall's Zapier integration, then engineered a custom event-correlation layer that captures both triggers, matches them reliably via the JustCall REST API, and pushes clean, accurate updates straight into the correct pipeline stage in LACRM — with a parallel sync to Google Sheets for reporting.
Result: zero manual data entry, real-time CRM records the sales team could finally trust, and full documentation so their internal team can maintain the system without ongoing developer support.
This is the kind of API integration and workflow debugging I do daily — across Make, n8n, Zapier, Python, and REST API/webhook architecture. If your CRM, telephony, or sales tools aren't talking to each other properly, let's fix that.::
1
76