The build

When something breaks, I open my editor, not a ticket.

None of this is a weekend side project. It's the stuff a real ad operation actually runs on, tracking, lead routing, research, reporting, built by the same person who spends the budget every day.

CAPIBuilt from scratch
90%Manual ops removed
6+AI coding tools daily
01 / Attribution

The Meta tracking, wired up by hand.

Beegru AI Boost needed conversion tracking that could survive iOS, ad blockers, and leads that come in offline. So I built the whole Conversions API pipeline myself, start to finish. No plugin, no agency.

01

Multi-pixel dispatch

One lead sends clean, de-duplicated events to every pixel at once, and quietly retries when Meta acts up.

  • Bulk event sending
  • Retry queue + dedup keys
  • Quality-scored events
02

Personal data handled properly

Every bit of personal data gets SHA-256 hashed before it leaves the sheet, so match quality stays high and no raw data ever goes out.

  • SHA-256 hashing
  • Full Site Dataset integration
  • Match quality 8+/10
03

Sub-second routing

The moment a lead lands, it's cleaned up, locked from edits, handed to the next sales rep in line, and pushed out over WhatsApp and email.

  • Round-robin RM assignment
  • Phone auto-formatting
  • WhatsApp + email notify
02 / Method

How a tool goes from annoying to done.

01 Hit the wall Something in a live campaign gets in my way, tracking breaks, research is all manual, a list is full of junk. in the account
02 Figure it out I work out what it needs to do against the real data and the real screen it has to fix. against real data
03 Build it Put it together fast with my AI coding setup, then patch the edge cases only someone who runs ads would catch. claude + codex
04 Ship + use it Put it live on a subdomain or Docker and actually use it on real spend, instead of letting it rot in a repo. in production
03 / The stack

Built fast, then hardened by someone who runs ads.

The actual building happens with an AI coding setup I use every day. But knowing what's really broken in a campaign, and when a thing is genuinely done, that comes from four years spent inside the ad account.

That mix is the whole point of Serves: the person writing the retry logic is the same person who watched the leads dry up.

AI coding workflow
Claude Code Antigravity GitHub Copilot Codex Ollama DeepSeek v3 Claude Pro + API Gemini
Runtime & automation
Google Apps Script Python · Flask Chrome Extension APIs WhatsApp Cloud API Meta Marketing API Docker n8n Vercel
Tracking & measurement
Meta Conversions API GA4 Google Tag Manager UTM architecture A/B testing
04 / Disciplines

Two jobs, one person, nothing lost in between.

Media buying

Getting the leads in: 6+ accounts, ₹25L+ a month, high-ticket real estate at under ₹1K a lead.

  • Meta, Google, LinkedIn
  • Creative testing discipline
  • Learning-phase analysis

Tracking & attribution

Keeping it honest: making sure every rupee of spend is measured right, end to end.

  • CAPI + server-side GTM
  • Event taxonomy
  • Dedup + match quality

Automation

Cutting the busywork: everything between a lead landing and a report going out, handled on its own.

  • Apps Script pipelines
  • Lead routing
  • Auto-reporting

Building the tools

Turning all of that into tools other advertisers can use too, not just me.

  • Web apps + extensions
  • Docker deploys
  • Live subdomains
Not your average marketing hire

A growth hire who can also build the plumbing.

If that's the piece missing on your team, the same person who wrote all this can build it for you.