Work

Systems that run every day, in real businesses.

Case studies in marketplace and ERP automation, AI agents, SEO systems, and the data infrastructure underneath.

SEO · Content systems

Took reprolegal.com out of Google's sandbox

Built the technical SEO and content system for our own site. It cleared the sandbox and started bringing in leads with zero ad spend.

150+ leads · zero ad spend
Automation · Marketplaces & ERP

End-to-end order automation across marketplaces and an ERP

Orders flow into the ERP, get fulfilled, and confirm back automatically — inventory synced across every channel. Works the same on NetSuite, SAP, Oracle, or Odoo.

Continuous · no manual entry
AI agent · Finance ops

Turn a chat message into an approved invoice

A bot reads a PDF, extracts the line items, and routes an ERP invoice for approval — no manual entry. Works with any chat tool or ERP.

One step · not five
Data engineering · Ad spend

Caught a doubled ad spend line and an $11B phantom revenue bug

A daily reconciler between ad accounts and the warehouse — caught doubled spend, a silently-failed export, and a currency bug inflating GA4 revenue by nine orders of magnitude.

$6.9k/day · double-counted spend caught
AI agent · Media buying

Automated campaign launches across dozens of countries

An agent that scans Drive and a spreadsheet and builds the full campaign — cards, ad groups, budgets, UTMs — paused, logged, one click from live.

Dozens of countries · one pipeline
Marketing analytics · Ad reporting

Turned day-old ad reporting into hourly

An hourly pipeline across Google Ads, Meta, and X, reconciled to the cent against daily numbers — plus a conversion-value bug fix.

30h → 1h · reporting lag, multi-brand
Data engineering · BigQuery

Cut a nightly BigQuery scan by 95%

Found the real cause of nightly pipeline timeouts — unpartitioned sources — and verified the fix value-for-value, not just row counts.

−95% · nightly scan, 0 timeouts
Data engineering · FinOps

Automated financial reporting from scratch

Automated ingestion and transformation into one warehouse with data-quality checks and real-time metrics — and cut cloud infra cost 40% along the way.

3d→0 · monthly reporting time
Marketing analytics · Ecommerce

Diagnosed a ROAS collapse, mapped the path to 400%

Disproved the fraud/geo theory with raw data, found the real cause, and rebuilt the conversion architecture around offline closes.

400% · blended ROAS target
Marketing analytics · Subscription

Fixed attribution, returned conversions to Google Ads

Payment domains were stealing credit from the real channel. We patched the attribution models and closed the loop back to Google Ads.

283/408 · sessions reattributed
AI bots · Lead ops

Intelligent bots bridging disconnected systems

Bots that sync data in real time, classify and qualify leads instantly, and resolve common requests from a knowledge base.

70% · comms automated
ML · Predictive

Predictive CAC — forecasting lead value

Trained a maturation model on historical lead-to-close data and fed predicted value into Smart Bidding.

180d · value maturation horizon