Product Analytics with GA4
GA4 data modeling across e-commerce and financial services clients using dbt and Dataform. Medallion architecture with specialized marts for user behavior and product interactions. FinOps optimization for cost and performance.
GA4
dbt
Dataform
BigQuery
KestraFinOps
Data StudioRaw GA4 exports were inconsistent and difficult to reuse across teams and industries.
Medallion architecture (bronze, silver, gold) with dbt and Dataform. Specialized marts for user behavior, product interactions, and simulations. FinOps strategies for cost optimization. Data Studio dashboards for analytics delivery.
Unified GA4 models enabling product insights and reliable KPI tracking. Data Studio dashboards delivered to business teams. Saves 6 hours per week in analytics delivery.
GA4 FinOps & Performance Optimization
Refactoring GA4 workloads in BigQuery: query optimization, partitioning, clustering, and governance implementation.
BigQueryGA4SQLFinOpsKestraGA4 queries were inefficient with high costs and inconsistent patterns requiring refactoring.
Refactored queries with partitioning, clustering, and query optimization. Implemented cost monitoring and governance guidelines.
Refactored GA4 workloads resulting in lower costs and improved query performance for analytics teams.
Marketing Analytics Platform Migration
Technical lead for migration from Dataroma to Modern Data Stack. Multi-source integration (Catchr, Couchdrop) and DSP data extraction. dbt standards implementation, platform documentation, and client team training.
dbtKestra
GCP
PythonCouchdropCatchrLegacy Dataroma infrastructure was brittle and required migration to a modern platform.
Migrated to Modern Data Stack. Technical lead for multi-source integration (Catchr, Couchdrop) and DSP extraction via Couchdrop. Built dbt standards, documented platform, and delivered client training.
Modern, scalable marketing analytics platform with standardized dbt practices and unified multi-source integration.
Multi-Store Retail Analytics & Monitoring
Refactoring and optimization of multi-store dashboard platform. Dashboard architecture redesign, Row-Level Security (RLS) implementation, consolidated global dashboard, and automated Slack alerting.
dbt
Slack APIData StudioGCPGA4Multi-store dashboards needed secure access, architecture refactoring, and monitoring for missing data.
Dashboard architecture redesign with Row-Level Security (RLS) for multi-store access. Built consolidated global dashboard and automated Slack alerting for missing store data.
Refactored dashboard platform with secure multi-store access, consolidated views, and automated monitoring.
Modern Data Foundation & Orchestration — 5 industries.
Medallion architecture with Kestra orchestration and ingestion (Airbyte, Stitch, Funnel). Modular dbt models with marts layer. Automated retry mechanisms, error handling, and data quality framework.
Kestra
AirbyteStitchFunneldbtBigQuery
SAPMultiple disparate data sources requiring consolidation. Time-consuming manual reporting and lack of confidence in numbers.
Medallion architecture (bronze, silver, gold) with Kestra orchestration. Airbyte for source integration. Modular dbt models with marts layer. Automated retries, error handling, and data quality testing.
More than 5 hours saved per week for reporting. 100% confidence restored in numbers. Comprehensive monitoring across net sales, bundle performance, upsell rates, and customer acquisition costs.
Data Monitoring — Slack-Based Analytics Alert System
Slack monitoring system for dbt pipelines and business KPIs. Automated alerts for failures and errors. Scheduled KPI reporting with thresholds. Interactive Slack commands for data access. Delivered across 5 industries.
Slack APIKestraBigQueryPythondbtData quality issues and dbt failures detected too late. Business teams lacked real-time KPI visibility.
Slack API integration with dbt and BigQuery. Automated alerts for test failures and pipeline errors. Scheduled business KPI reporting with thresholds. Interactive Slack commands for data access.
Reduced time to detect failures through instant Slack notifications. Scheduled KPI delivery in team channels. Improved collaboration and faster incident response.
Plein Data — Edge Analytics Lakehouse — Building a portable, vendor-agnostic analytics stack on GCP.
Redesigning a 20-year fuel price ecosystem on GCP with portability at its core. Daily ingestion feeds dbt transformations; a Python bridge offloads Parquet to Cloudflare R2, bypassing BigQuery egress. Analytics served via DuckDB WASM at the edge — full compute-storage decoupling.
dbt fusion
Cloudflare R2
DuckDB WASMPython
TerraformA 20-year-old fuel price dataset on GCP/BigQuery with growing egress costs and full vendor dependency — needed a portable, cost-zero architecture.
Built on GCP with portability by design. Daily ingestion feeds dbt fusion transformations. A Python bridge offloads Parquet files to Cloudflare R2, bypassing BigQuery egress. Analytics served via DuckDB WASM at the edge — full compute-storage decoupling with Terraform-managed infra.
Targeting 0€ egress cost, 100% portable logic across any cloud provider, and zero vendor lock-in. Full Apache Iceberg compatibility for open table format.
StackRadar — Data Stack Release Watcher
Your data stack evolves in silence — StackRadar listens for it. Developed with Specification-Driven Development. A daily Cloud Run job fetches GitHub releases across dbt, Airflow, BigQuery, and more, analyzes them with mistral-small-latest, and delivers a structured email digest every morning.
PythonCloud RunGCPMistral AICloud SchedulerKeeping up with majors and minors across data tools meant juggling LinkedIn posts, GitHub changelogs, and newsletters — and still discovering breaking changes too late, or missing long-awaited features entirely.
Built with Specification-Driven Development: full spec and acceptance criteria before writing a single line of code. A Cloud Run Job fetches GitHub releases each morning, analyzes them with mistral-small-latest (Mistral AI), and sends a structured digest by email. Monitors dbt and its packages, Airflow, BigQuery, Lakehouse, and other data tools.
Open source and in production. Release monitoring fully on autopilot — breaking changes and new features land in the inbox, not in a post-incident report. Readable, maintainable, and extensible codebase thanks to the SDD approach.
Berlioz — AI Nutrition Assistant for Cats
RAG chatbot helping cat owners evaluate pet food for renal and urinary conditions. Semantic search over a veterinary knowledge base with URL analysis, OCR label reading, and an AI expert chat.
OpenAIUpstash Vector
Node.jsVanilla JSCat owners managing renal disease or urinary conditions had no reliable tool to evaluate pet food labels against evidence-based veterinary thresholds (phosphorus, proteins, Ca/P ratio, acidifiers).
RAG-powered web app with Vanilla JS frontend and Node.js serverless API. Upstash Vector for semantic search, OpenAI embeddings, Groq LLaMA for inference. URL-based product analysis and OCR label reading over a curated veterinary knowledge base.
End-to-end product deployed on Vercel with infra cost < $0.15/month and <1.1s response time. Semantic RAG search grounded in evidence-based thresholds — 0 hallucination risk on core nutritional facts.
Automated Tech Watch System
Automated content aggregation and analysis from email and RSS sources to build a continuous intelligence pipeline.
PythondbtBigQueryCloud RunTerraformManual tech watch was time-consuming and inconsistent across sources and formats.
Python ingestion for email and RSS feeds, dbt transformations, dlt loading to BigQuery, and serverless deployment on Cloud Run with Terraform-managed infra.
Automated knowledge base centralizing technical monitoring. Saves 5 hours per week in research time with consistent coverage across all sources.