What I've Built
Client Work & Internal Tools

Imprint Digital

Marketing Agency • 10+ Active Clients • Growing Team

The Problem:

Agency tracking 42 different KPIs across Stripe, Google Ads, Facebook, Go High Level (CRM), and ClickUp. Team spent 15+ hours every week manually copying data into spreadsheets for client meetings and internal reviews.

What I Built:

• Custom Google Sheets system tracking all 42 KPIs with automated rollups and status indicators
• 6-page Looker Studio dashboard (Overview, Sales, Marketing, Production, Client Success, Finance)
• One-click monthly reporting with automated data snapshots via Google Apps Script
• Department-specific views so each team sees only their metrics

ESB Advertising

National Digital Marketing Agency

The Problem:

Data scattered across Google Sheets with manual updates. No centralized source of truth. Analysts spending hours updating dashboards manually. Leadership couldn't trust the data.

What I Built:

• BigQuery data warehouse - Centralized repository for all business data
• API integrations - Cloud Run services pulling data from all platforms automatically
• Looker Enterprise dashboards - Real-time visibility for execs and team leads across 15+ client accounts
• Automated data refresh - Hourly updates, zero manual intervention

ESB Advertising

Multi-Platform Data Collection Bot

The Problem:

Multiple advertising platforms don't provide API access. Analysts manually logging in daily to download CSVs, transforming data in Excel, then uploading to BigQuery. Error-prone and time-consuming.

What I Built:

• Selenium automation script - Logs into platforms automatically with credential management
• Data transformation pipeline - Standardizes CSV formats for BigQuery schema
• Cloud Storage integration - Saves historical CSVs for audit trail
• Daily Cloud Run job - Runs on schedule, zero manual intervention

GOWD

Privacy Law Compliance Monitoring

The Problem:

Law firm needed to analyze thousands of websites for California privacy law violations. Manually checking cookie consent banners and reject buttons was impossible at scale.

What I Built:

• Automated web scraper - Playwright bot visits websites using California VPN
• LLM vision analysis - AI analyzes screenshots to detect cookie banners and reject buttons
• Compliance detection - Identifies one-click reject compliance vs multi-step violations
• Google Sheets export - Automated QA results with evidence screenshots

Reddit Lead Intelligence System

Proposed for Wealth Management Firm

The Challenge:

Wealth management firm wanted to find high-value leads organically on Reddit. Manual monitoring was impossible—thousands of conversations happening daily across 15+ subreddits.

What I Built:

• Real-time Reddit monitoring - Scans 10-15 subreddits every 1-2 hours for relevant conversations
• AI opportunity scoring - LLM analyzes each post/comment, scores 0-100 based on lead quality
• Profile vetting system - Filters out bots, spam accounts, and low-quality prospects
• PostgreSQL tracking - Logs all opportunities with context, reasoning, suggested responses

AI Data Assistant: Ask Your Data

National Mattress Retailer

The Challenge:

Business executives needed quick access to marketing data insights but required technical teams to write SQL queries for every ad-hoc request. This created bottlenecks, delayed decision-making, and consumed significant engineering resources for routine data questions.

What I Built:

• Natural language interface - Ask questions in plain English, get instant BigQuery insights
• OpenAI Assistants API - Vector embeddings for schema knowledge, automatic SQL generation
• Secure BigQuery bridge - Read-only access with SQL injection prevention and guardrails
• Production FastAPI backend - Deployed on Cloud Run with conversation persistence in Firestore

Automated SEO Audit Generator

Web Design & SEO Agency

The Challenge:

Agency needed to scale lead generation by providing value upfront through SEO audits, but manual analysis took 45+ minutes per prospect. Checking SSL, meta tags, mobile responsiveness, analytics, accessibility, and content optimization made it impossible to offer free audits at scale.

What I Built:

• 39-signal SEO extraction - Technical, on-page, analytics, CRO, mobile, accessibility analysis
• Dual-viewport crawling - Desktop (1920x1080) + mobile (390x844) screenshots with Playwright
• 5-score calculation system - Health, opportunity, CRO, technical, and quick-win scoring
• Professional HTML reports - Color-coded issues, visual evidence, prioritized recommendations

AI-Powered Legal Document Analysis

Privacy-First Document Intelligence • Property Dispute Resolution

The Challenge:

Property damage dispute requiring analysis of 1,000+ pages across 20 PDFs of condo governing documents to find maintenance responsibility clauses. Documents contained sensitive financial and personal information that couldn't be uploaded to external AI services. Manual review would take 15-20 hours or cost $3,000-5,000 in legal fees.

What I Built:

• Local AI extraction - Ollama vision model processing PDFs entirely on local machine (zero external API calls)
• Automated cleanup - Removed 648 code blocks and 9 duplicate sections while maintaining document integrity
• Intelligent search - 25+ legal/technical keywords across all documents, extracted 1,073 relevant sections
• 5-layer evidence framework - Built comprehensive legal argument with 95% confidence in successful claim