Data Infrastructure & Growth Architecture

Conversion Funnel & Bulletproof Tracking Architect

I engineer data pipelines, recovery infrastructures, and automated systems to solve attribution loss and creative fatigue. Translating complex technical systems into predictable marketing revenue.

Explore 6 Case Studies

Verified Credentials & Certifications

Azure AI Certified (AI-900)

Google Analytics 4

Google Ads Specialist

Attribution Telemetry Pipeline

System Architecture

The Invisible Factory: iOS 14+ Bypass & First-Party Telemetry

Standard portfolios focus only on static creatives. High-performance B2B brands prioritize telemetry. Bypassing browser-side adblockers and cookie restrictions, my server-side framework captures conversion events directly via secure container gateways, recovering 30%+ of otherwise lost attribution signals.

CONVERSIONS API

Match Score: 8.5/10

Clean user parameters mapped directly to Meta Ads endpoints.

DATA RECOVERY

+30% Signal Accuracy

Events recovered from browser-side blocks and browser constraints.

COOKIE LIFETIME

ITP 7+ Day Resiliency

Extends ephemeral browser cookies back to standard marketing windows.

PRIVACY ROUTING

GDPR/LGPD Anonymizer

Hashes personal customer details before database ingestion.

gtm_server_gateway.log

CLIENT BROWSER

ITP BLOCKED (30%)

CONTAINER GATEWAY

1ST-PARTY DATA

CONVERSIONS API (META/GA4)

DEDUPLICATED API

STATUS: ROUTING ENABLED SECURED FIRST PARTY

B2B Growth Engineering Cases

Click tabs on each card to inspect the Technical Specs and Source Blueprints.

6 Active Systems Displayed

Case 01 / Creative Engineering

Video Automation & Upscaling

$0.03 CPV

To bypass ad platforms' watermark tracking and creative exhaustion, I programmed a Python batch-rendering pipeline. It removes client watermarks, overrides chroma-key backgrounds automatically, and upscales low-res formats to 4K using bicubic filters. Scaled custom creative runs down to $0.03 CPV.

planta_baixa.png

dashboard_performance.png

# FFmpeg 4K Bicubic Scaler & Watermark Stripper
ffmpeg -i input.mp4 -vf "
  scale=3840:2160:flags=bicubic,
  delogo=x=10:y=10:w=120:h=40:show=0
" -c:v libx264 -crf 18 -preset slow output_4k.mp4

STATUS: OPERATIONAL

Case 02 / Telemetry Diagnostic

Affiliate Conversion Fix

$0.75 CPL

Glothub's Korean affiliate pipeline suffered from 'hallucinated conversions' optimization due to off-domain checkouts. I mapped checkout form fields into a custom dataLayer push and implemented retention triggers in GTM (minimum 30-sec scroll, outbound click filters) to rescue client pixel accuracy, securing leads at $0.75.

// User Data Injection for Enhanced Conversions
gtag('set', 'user_data', {
  'email': '{{USER_EMAIL}}',
  'phone_number': '{{USER_PHONE}}'
});
// Custom Telemetry Event Push
window.dataLayer = window.dataLayer || [];
window.dataLayer.push({
  'event': 'warm_lead',
  'time_on_page': 30
});

STATUS: OPERATIONAL

Case 03 / Data Engineering

Autonomous Data Pipeline

10h to 4m

I built a data extraction engine in Python to structure fragmented LMS contents. By coupling Playwright browser contexts via Chrome DevTools Protocol (CDP), the script intercepted runtime JSON quiz APIs, downloaded 75+ docx attachments, and aggregated everything into unified semantic Markdown formats ready for NotebookLM.

1. Playwright Init [CONNECTED_CDP]

2. Network Listener [JSON_INTERCEPTED]

3. Docx Parser Worker [75_FILES_MERGED]

4. Target Output [135KB_MD_READY]

# Capturing JSON Quiz responses dynamic network intercept
page.on("response", lambda response: 
  handle_quiz_json(response) 
  if "/api/v1/questions" in response.url 
  else None
)

STATUS: OPERATIONAL

Case 04 / Automation Agent

Agentic Vision AI: Extreme Automation of Protected LMS Systems

95% Success

I built an agentic vision system designed to audit complex corporate LMS platforms and extract secure data where traditional APIs fail. Connected to secure browser sessions via CDP (port 9222), it pierces nested iframes and handles shadow DOMs. If elements are obfuscated or DOM-blocked, the agent captures screenshots and queries multimodal models (Gemini Flash) to parse, structure, and audit interface data in real-time.

SOLVE SPEED < 1.5s / q

ACCURACY RATE 95% avg

# Gemini Vision Fallback block for iframe blocks
page.screenshot(path="exam_screenshot.png")
with open("exam_screenshot.png", "rb") as f:
    b64_image = base64.b64encode(f.read()).decode()
response = gemini.generate_content({
    "inline_data": {"mime_type": "image/png", "data": b64_image},
    "text": "Analyze screen layout, extract structured interface text and return target index."
})

STATUS: OPERATIONAL Discuss LMS Auditing

Case 05 / Outbound Lead Machine

Maps Scraper & Data Leak Auditor

100% Automated

I built a B2B prospecting tool (`scraper.py`) that crawls Google Maps places within S-Tier high-ticket niches. The script checks site contents to audit if the business is running browser-side Meta tracking pixels but lacks Server-Side Conversions API (CAPI) markers. Identified leaks export to CSV to feed an outbound cadences flow in n8n and Apollo.io.

[FINDING] Med Spa Los Angeles... website found.

[AUDIT] Pixel: YES | Server-Side: NO

[DATA LEAK DETECTED!] Saving to leads_qualificados.csv

# Data Leak Detector (Checking Browser Pixel without Server clues)
has_pixel = bool(re.search(r'fbevents\.js|fbq\(', html, re.I))
has_capi = bool(re.search(r'googletagmanager\.com/gtm\.js|fb_capis', html, re.I))
leaking_data = has_pixel and not has_capi
if leaking_data:
    results.append({'Name': name, 'Leaking': True})

STATUS: OPERATIONAL

Case 06 / Market Opportunities

Indeed & Workana Sniper Radars

Elite Filters

I built two specialized radars for automated job hunting: `workana_sniper.py` filters projects with budgets >= 500 BRL and < 15 active bids in GTM/CAPI categories to secure low-competition elite contracts. `indeed_farmer.py` coordinates Playwright CDP sessions to monitor and target easy-apply SDR/Tech vacancies, routing tailored pitches.

Workana filter threshold: >= 500 BRL, <= 15 Bids

Indeed active keywords: SDR, Analyst, GTM, CAPI

Response pitch routing: Microsoft/Google credentials pitch

# Selenium Workana sniper elite targets parser
if budget_value >= 500.0 and bids_count <= 15:
    elite_targets.append({
        'title': project_title,
        'budget': budget_str,
        'link': project_link
    })

STATUS: OPERATIONAL

Conversion Funnel & Bulletproof Tracking Architect

Attribution Telemetry Pipeline

The Invisible Factory: iOS 14+ Bypass & First-Party Telemetry

B2B Growth Engineering Cases

Video Automation & Upscaling

Affiliate Conversion Fix

Autonomous Data Pipeline

Agentic Vision AI: Extreme Automation of Protected LMS Systems

Maps Scraper & Data Leak Auditor

Indeed & Workana Sniper Radars

Engineering Certifications Directory

What is your business's monthly revenue?

How much do you spend on ads monthly?

What is your core tracking or funnel bottleneck?

Submit contact details to generate diagnostic

Telemetry Analysis Completed