Analytics Engineering · Data Engineering · AI Harness Engineering
I'm Gonzalo Brunoldi, an analytics engineer with 6+ years building the pipelines, models and dashboards that turn messy sources into numbers teams trust — and the AI harnesses that let agents do that work reliably.
01 / about
I build the data infrastructure that teams actually trust: pipelines that land raw sources, models that turn them into clean, governed tables, and dashboards people can answer their own questions with.
Since 2020 I've moved from analyst to data team lead to engineer — across health-tech, e-commerce, fintech and collectibles — and I keep the business question in the room while I work on the plumbing. Lately that includes applying AI to the workflow itself.
02 / impact
- 3h → 25mMarketing pipeline runtime after a full rebuild at IPSY — with more metrics, not fewer.
- 1stAttribution model in IPSY's marketing area, linking communications to orders and subscriptions.
- SF → DBXLed the migration of multiple pipelines from Snowflake SQL to Databricks / PySpark at Clip.
- E2EOwns reporting end to end at Fanatics: AWS DMS → dbt silver & gold → semantic models → Sigma.
03 / ai harness
I don't just use AI. I engineer the harness around it.
A model is only as good as the context, tools and guardrails it runs inside. At Fanatics I led AI tooling adoption for the data team; on my own time I run a multi-agent setup that plans, builds, reviews and documents its own work — and I treat it like any other production system.
harness:
model: claude
context: [project docs, playbooks, memory] # reusable, versioned
skills: 70+ workflows # plan, ship, review, debug
playbooks: 30 failure recipes # hard-won, never relearned
hooks: token-optimized guards # cost down, signal up
tools: mcp: [browser, db, tracker, deploy]
agents: planner → implementer → qa # parallel, gated
verify: [tests, types, lint] before merge
- 01
Context engineering
Reusable context docs and cross-project playbooks so agents start informed, not from zero — the same discipline as a well-documented semantic layer.
- 02
Agent orchestration
Planner, implementer and QA agents working in isolated worktrees with acceptance tests as the gate — fan-out where it pays, a single agent where it doesn't.
- 03
Guardrails & cost
Hooks that trim tokens and block unsafe actions, plus verification before anything merges. Faster delivery, lower spend, fewer surprises.
04 / experience
- 2025 — now
Analytics Engineer
Fanatics Collectibles
End-to-end reporting on a medallion architecture; dashboarding framework overhaul with lineage and governance guardrails; reusable A/B experiment tracking; led AI tooling adoption.
- 2024 — 2025
Data Engineer
Clip · Financial Services
Scalable ETL with PySpark, AWS Glue and Databricks; Airflow DAGs for ingestion and quality checks; Snowflake → Databricks migration that cut cost.
- 2022 — 2024
Senior Data Analyst, Marketing · Data Analyst, Web
IPSY
Sole owner of marketing analytics; first attribution model; pipeline from 3h+ to 25 min; master tables as single source of truth for KPIs and e-commerce funnel analysis.
- 2020 — 2022
Data Team Lead · Data Analyst
¡Appa!
Set the data team roadmap with the CTO; built a self-service culture through SQL, Python and dashboard workshops; real-time BI on PostgreSQL; clustering models.
- 2020 — 2021
Data Analyst · Assistant Professor
ÜMA Health AI · Coderhouse
KPIs defined directly with the CEO from Firebase and BigQuery data; taught SQL, data modeling, Power BI and storytelling.
- 2017 — 2019
Strategy & Business Consultant
MealPal (Sydney) · Accenture
Split-testing and C-level insights at a startup; SAP implementation and process optimization projects for MetroGAS and Santander Río.
05 / stack
Engineering
- Python
- SQL
- PySpark
- dbt
- Apache Airflow
- AWS Glue
- AWS DMS
- Medallion architecture
Platforms
- Databricks
- Snowflake
- BigQuery
- PostgreSQL
Analytics & BI
- Sigma
- Tableau
- Power BI
- A/B testing
- Attribution modeling
AI & Automation
- Claude / Claude Code
- AI harness engineering
- Context engineering
- Multi-agent orchestration
- MCP
- Workflow automation
06 / contact
Have data that should be telling you more?
Let's talk.