DATA ENGINEERING · AI ENGINEERING · CROSS-INDUSTRY

NapoliData

17 years in data. 5 industries, 7 employers, no gaps.

Senior Data & AI Engineer. Independent since March 2026. Verify against LinkedIn in 30 seconds.

See experience → LinkedIn → Let's talk
// trajectory
2026 → NapoliData · Independent
2024 — 2026 · Proactiviti (Data & AI)
2023 — 2024 · Fivvy (Data analytics)
2022 — 2023 · Aprende (BI)
2020 — 2021 · Prisma Medios de Pago
2016 — 2020 · Banco Galicia & Galicia Seguros
2009 — 2016 · Banco Patagonia
17 years · 5 industries

The trajectory.

Seven employers, in reverse order. Every entry maps to a LinkedIn role and a stack you can interrogate on a call.
MAR 2026 — PRESENT
NapoliData · Independent
Senior Data & AI Engineer. Distributed data pipelines and cloud platforms across AWS and Azure. LLM-based observability and AI assistants (LangChain, Claude, Amazon Bedrock, Ollama). Stack: Python, Spark, Databricks, Snowflake, Kafka, Airflow on Kubernetes, dbt.
2024 — 2026
Proactiviti
Data Engineer. Distributed pipelines on AWS (Glue, Lambda, Step Functions, Athena) and Airflow on Kubernetes. Hybrid flows across Azure Data Factory, Synapse, and Databricks. −20% unplanned incidents through LLM-based monitoring agents. Tuned Redshift and PostgreSQL workloads.
2023 — 2024
Fivvy
AWS Data Engineer. Modular ETL pipelines using Lambda, EMR, Glue, Step Functions, orchestrated with Airflow on Kubernetes. −30% S3 storage costs via lifecycle and partitioning. Tuned Redshift for analytical workloads.
2022 — 2023
Aprende Institute
AWS BI Data Engineer. Cloud-native ETL with Glue, Lambda, Step Functions, Redshift. −35% processing costs · −50% manual reporting · +20% marketing ROI via predictive models, QuickSight automation, and centralized data lake.
2020 — 2021
Prisma Medios de Pago
Data Scientist Project Leader. Led Big Data & Analytics projects for Argentina's largest payment processing network. Stack: S3-Athena, Python, PySpark, Docker.
2016 — 2020
Banco Galicia & Galicia Seguros
Data Analyst → Senior Data Scientist (Marketing & BI). Built propensity, segmentation, and churn models for Galicia Seguros across life, home, and auto insurance products. In parallel, ran customer analytics for banking products (loans, cross-selling) using RFM and unsupervised techniques, plus NLP for re-marketing chatbots. Measurable lift in sales and material reduction in call-center costs.
Case study overview →
2009 — 2016
Banco Patagonia
Data Analyst (Credit Risk Management). Developed and maintained credit scoring models for retail and corporate clients across six years. Stack: Python, SPSS, SQL.

How I work.

Three recurring shapes of project. Each one: where the data comes from, what gets built, what ships.
01 · OBSERVABILITY · LLM AGENTS
From log noise to actionable incidents.
−20% unplanned incidents
Proactiviti — verifiable on CV
Python LangChain Bedrock / Claude Airflow on K8s CloudWatch
▸ Input
Airflow task logs · CloudWatch alarms · dbt test failures · Slack noise from on-call channels.
▸ Build
LLM classification agent over normalized events. Correlation across sources, severity scoring, runbook retrieval via embeddings.
▸ Output
Triaged incident notification with root-cause hint, linked runbook and suggested next action. Fewer pings, faster MTTR.
02 · DATA ENGINEERING · CLOUD
Heterogeneous sources into a queryable warehouse.
−30% S3 · −35% processing
Fivvy & Aprende — verifiable on CV
AWS Glue Step Functions Databricks Snowflake dbt Kafka Azure ADF / Synapse
▸ Input
Postgres · S3 dumps · SaaS APIs · Kafka streams. Mixed schemas, mixed cadences, no shared dictionary.
▸ Build
Modular ETL/ELT on AWS (Glue, Lambda, Step Functions) or Azure (ADF, Synapse). Spark on EMR/Databricks for heavy load. dbt for modeling and tests. Cost & partitioning tuned from day one.
▸ Output
Curated tables in Redshift / Snowflake. BI-ready, SLA-tracked, documented. Storage and compute costs cut by a third.
03 · PREDICTIVE MODELING · MLOps
From raw transactions to scored populations.
+20% marketing ROI
Aprende — verifiable on CV · 10 yrs banking & insurance
Python scikit-learn MLflow Embeddings + RAG QuickSight / Power BI
▸ Input
Transactional warehouse · CRM attributes · behavioral signals · campaign history.
▸ Build
Feature engineering + classical models (logistic, GBM, RFM, unsupervised) or embeddings & RAG when text dominates. MLflow tracking, batch inference in production.
▸ Output
Scored customers piped to CRM, campaigns, credit decisions, churn watchlists. Decisions move from gut to evidence.

Stack I connect.

Not a laundry list. Tools I've shipped to production, grouped by where they fit in the data path.

Cloud · Pipelines

AWS Glue Step Functions Lambda EMR Athena Redshift Azure ADF Synapse Airflow on K8s Kafka

Processing · Modeling

Python PySpark Databricks Snowflake dbt Delta Lake Docker

AI · LLMs · ML

Anthropic Claude Amazon Bedrock LangChain Ollama Hugging Face scikit-learn TensorFlow / Keras MLflow

BI · Delivery

QuickSight Power BI Tableau Metabase

Highlighted = primary, used in production within the last 2 years. Rest = working knowledge, deployed in earlier roles.

What I write, I keep public.

Production-grade code in public repositories. Review it before you hire me.
Databricks · lakehouse patterns
4 repos →
DATABRICKS · ETL
Medallion Pipeline
Bronze → Silver → Gold with Delta Lake, PySpark, data quality and optimization.
View repo →
DATABRICKS · DLT
Delta Live Tables
Declarative pipelines with CDC, streaming, and built-in data quality.
View repo →
DATABRICKS · ML
MLflow end-to-end
Feature engineering → training → registry → batch inference in production.
View repo →
DATABRICKS · PERFORMANCE
Cost optimization
Skew handling, partitioning, caching, cost management on production clusters.
View repo →
dbt · warehouse-native analytics engineering
4 repos →
DBT · MEDALLION
Staging → Intermediate → Marts
Sources, contracts, snapshots, custom macros, exposures. Full warehouse-native medallion.
View repo →
DBT · TESTING
Data quality, layered
Generic + singular tests, dbt-expectations, unit tests (1.8+), Elementary patterns.
View repo →
DBT · INCREMENTAL
Four strategies, side by side
merge · insert_overwrite · append · microbatch (1.9). Backfill safety and late-arriving data.
View repo →
DBT · PERFORMANCE
Cost & performance
Materialization choice, clustering, query tagging, warehouse spend monitoring from metadata.
View repo →
View all repos on GitHub →

Let's talk.

Three modes of working together:

PROJECT-BASED
Defined scope and timeline. Typically 4–12 weeks.
FRACTIONAL
Senior part-time capacity, embedded in your team.
ADVISORY
Architecture, code review, technical decisions.

Also available for adjacent verticals — payments, e-commerce data, logistics, and other data-heavy industries.

A 30-minute call to see if there's a fit. No pitch.

LinkedIn WhatsApp