reasoning_bytes
Senior Data Scientist & Systems Architect

Diego Torres Dho Architecting Scalable AI

Bridging theoretical modeling and robust production deployment — Financial AI, LLM strategy, and mission-critical infrastructure.

View work
9+
Years in AI / ML
~24
Years in Systems
4
Patents
career_arc.json // 1996 → now drag a node ·

Core Technologies

Python PySpark Docker / K8s LLMs & RAG Oracle DB CI/CD · MLOps

01About

Two worlds, one architect.

I'm a Senior Data Scientist and Systems Architect with over 8 years of specialized AI/ML experience, built on a 25-year foundation in mission-critical infrastructure.

My career has been defined by bridging two worlds: the theoretical complexity of machine-learning models and the rigorous demands of production engineering. I don't just build models — I architect scalable solutions to hard problems in Financial AI and predictive analytics.

02Experience

The career arc

Senior Principal Data Scientist

2024 — 2026

Oracle NetSuite AI

  • Shipped invoice payment-date forecasting model to production (XGBoost, MLflow) across 42,000+ enterprise accounts — +8% cash-flow prediction accuracy, materially less manual finance review.
  • Built predictive duration models for task and project scheduling — cut planning estimation error ~13% inside the NetSuite PM module.
  • Architected end-to-end anomaly detection for financial transactions — ~68% recall at <5% false-positive rate, directly reducing customer risk exposure.
  • Selected as 1 of ~35 engineers company-wide for the Oracle AI safety initiative; evaluated 15+ AI systems pre-release.
  • Delivered two keynote sessions at SuiteWorld 2024 (Las Vegas): "Streamline the Period Close: Harnessing AI and ML to Improve Efficiency" and "Scaling ML Pipelines for Financial Exception Management."

Senior Data Scientist

2021 — 2024

Oracle NetSuite AI

  • Fine-tuned open-source LLMs (T5-Large with LoRA) for internal document Q&A and text-enhancement PoC — directly informed Oracle's investment decisions on LLM-based product features.
  • Built bank-record reconciliation models to automate bank-to-general-ledger transaction matching for enterprise customers.
  • Led cross-functional teams on high-impact AI initiatives; presented AI/ML solutions and roadmaps to C-level executives.
  • Delivered technical session at SuiteWorld 2023: "Anomaly Detection Techniques with Business Use Cases".

Senior Data Engineer

2020 — 2021

Glovo · Marketing and Growth

  • Built end-to-end PySpark + S3 + Redshift pipelines from raw ingestion to analytics-ready consumption for a new internal system.
  • Replaced a legacy data provider via a new external-integration pipeline — saved >€150K/month (~€1.8M/year) in vendor costs.
  • Built an asynchronous cohort-reporting system integrating external sources with the data lake — enabled sharper marketing targeting and measurable churn reduction.
  • Implemented data-quality monitoring across three solution endpoints to detect sync failures, improving reliability of business-critical reports.

Early Data Science Roles

2017 — 2019

JPMorgan Chase · Kimberly-Clark · Datastar · Freelance

  • JPMorgan Chase (Sr Data Scientist) — Built a predictive ML tool (linear regression, SVM, decision trees, random forests) to forecast additional resource needs during scheduled maintenance windows — targeted ~5% improvement in resolution times.
  • Kimberly-Clark (Sr Data Scientist) — Engineered a predictive-maintenance model (LSTM, ARIMA) on Azure Databricks forecasting chemical spills in U.S. production lines (~$1M/year savings per line); built an RFM + K-Means customer-segmentation product for Global Marketing (~2% annual revenue uplift); prototyped an NLP CV-ranking tool on Azure for Global HR.
  • Datastar (DS & Analytics Specialist) — Engineered high-volume Sqoop ETL migrating 1.2B rows from Oracle to Hive in under 4 hours; built supervised/unsupervised R models for pharmaceutical clients.
  • Freelance (DS Consultant) — Built predictive churn models in SparkML on Hadoop for a major telecom provider; architected data models and ETL workflows for the data-warehouse-to-production-model transition.

↓ Infrastructure & Systems Architecture Era

Senior Principal Software Engineer

2015 — 2017

Oracle

Led LATAM performance engineering; delivered 200%+ resource-utilization gains and processor-architecture workshops.

Systems Principal Sales Consultant

2013 — 2014

Oracle

Architected SPARC / Engineered Systems solutions for C-level customers; contributed to $15M revenue target.

Managing Partner

2006 — 2013

IT evoluxion

Built a Sun/Oracle/Red Hat services firm; team of 11 engineers, 25+ enterprise clients.

Early Career (Infrastructure & High Availability)

Sun Microsystems, Banco Santander

Mission-critical Unix (Solaris/AIX/HP-UX), HA clusters, DR and storage for major finance and utility customers.

03Research & Intellectual Property

Patents & publications

Journal Publication Nov 2024

Machine learning for anxiety and depression profiling and risk assessment in the aftermath of an emergency

Artificial Intelligence in Medicine · Volume 157

Published Patent

Point-in-Time Relative Outlier Detection

Issued Patent

Generating Enhanced Queries Using Machine Learning Models

Filed Patent

Generating Vector-Based Recommendations

Published Patent

Cross-Cluster Transaction Risk Assessment

04Selected Work

From production systems to the page.

A few representative deep-dives — case studies and talks, not throwaway demos.

Case Study

Financial Anomaly Detection

A scalable pipeline screening 10M+ daily transactions for relative outliers with unsupervised learning and distributed compute.

PythonPySparkIsolation Forest
Conference Talk

Scaling ML Pipelines

SuiteWorld 2024 keynote on moving financial exception management from legacy PL/SQL to a distributed PySpark architecture.

KeynotePySparkOCI
Read the field notes & talks

05Contact

Open to senior AI/ML roles & select consulting.

After shipping production AI at Oracle NetSuite, I'm exploring what's next — open to senior data-science roles and a few consulting or architecture-review engagements. If you're moving from POC to production or scaling high-load ML pipelines, let's talk.

diego.torres.dho@gmail.com
Connect on LinkedIn