Pablo G. Baretto

Solutions architect and data engineer with 20+ years of experience. I design and build AI stacks: production systems, evaluation, search, agentic harnesses, and the data platforms under them. I design around each company’s cloud, vendors, privacy rules, cost structure, and the team that will run the system. That includes private and air-gapped deployments when shared model APIs are not acceptable.

I previously worked as a presales and delivery architect for production cloud and data platforms. The work starts from operating constraints.

Experience

AI architecture and implementation

2021 - Present

Long engagements with mid-market companies, product companies, and venture-backed startups. I assess the environment, design an AI stack that fits it, and implement the system far enough that the client can operate it.

US mental-health provider (regulated reporting). Built a generative workflow for clinical reports with evaluation against real cases and human review for exceptions, using existing data and process rather than a new model stack. About 98% of reports passed the agreed checks; roughly two reports per day still go to review.
Legal SaaS, Spain. Answers grounded in national law, with no cleaned corpus and no custom LLM. Ingestion pipelines plus RAG with citations. Agreed accuracy on sampled answers ~99%; human review dropped to about 1% of responses; time to an answer fell ~90%.
Matchbook AI, USA. Matching engine for messy, continuously arriving master data used by ~50 S&P 500 clients. Initial load ~500 million records and ~1 million updates per day (about 15% above the original target). Processing time down 35%; SLA breaches stopped.
Sonarplatform, EU. Search and analytics over on-chain data: ~1 billion transactions across 10+ Ethereum networks, ingest ~3,000 tx/s, API p95 under 800 ms.
US consumer-data product. Hybrid lexical and vector search under ingest and latency limits (ingest up to 200 MB/s, search under 500 ms), rather than a single-vendor search stack.
Holobrain is a private deployment pattern used when customer data and models must stay on the customer’s infrastructure. It’s one implementation path, not the architecture for every client. holobrain.ai

Data Analytics Solutions Architect, Logicalis (Spain)

2015 - 2021

Progressed from Cloud Solutions Architect to Data Analytics Solutions Architect at a global IT services firm. Designed production data and analytics platforms from each client’s cloud footprint, tooling, and operating model, including presales architecture and delivery. Not a standard reference build.

Architected and led Orange’s unified log and metrics platform on an Elasticsearch-inspired stack already compatible with their operations model. Reduced MTTD ~55%, MTTR 45%, unplanned downtime 30-40%, and manual log analysis >60%.
Delivered production systems and PoCs for Vodafone, Telefónica, FNMT, Grupo Piñero, La Caixa, Ocaso, AENA, and Catalana Occidente, scoped to each organization’s capacity and vendor landscape.

Senior Data Analytics Consultant

2006 - 2015

Analytics for IT monitoring and support across Latin America. Assembled open-source and existing vendor tools to fit mid-sized firms’ budgets and operations teams.

FacilDesk (founder). Open-source IT monitoring (Elasticsearch) and ticketing (OTRS) with analytics for mid-sized firms. Mitsubishi Motor Company: +67% customer satisfaction. Bancaribe: 43% faster incident response. Also Agroamerica and Cerámicas Caribe.
Sun Microsystems contractor (Venezuela). Data-centric monitoring assembled onto telco and oil & gas estates (CANTV, Movilnet, Telefónica, PDVSA, ENI).

Java Trainer and Consultant, Sun Microsystems (Venezuela)

2001 - 2006

Java training and consulting across northern Latin America.

Selected work

Holobrain. Private, air-gapped AI deployments for organizations whose privacy requirements rule out shared LLM APIs. Open-weights models and automated workflows run on the customer’s infrastructure. This is one implementation path when policy requires it, not the default for every client. holobrain.ai

Certifications

Google Cloud Professional Data Engineer, 2020

Elasticsearch Certified Engineer, 2020

Education

Universidad de Los Andes (Venezuela), Systems Engineer, 2001

Universidad Complutense de Madrid (Spain), Master’s in Marketing Management