Alejandro – Databricks, Python, SQL

An AI-powered compliance review solution for large pharmaceutical companies, where every piece of promotional and medical content must pass Medical, Legal, and Regulatory (MLR) review before it can be released. Manual MLR review cycles are long and error-prone, delaying time-to-market for marketing and medical materials. The solution embeds AI review assistants into the content approval workflow that check materials against compliance rules and flag issues for human reviewers. It is built as a scalable, productized offering that can be deployed across multiple enterprise pharma clients.

• Built AI-powered review assistants for Medical, Legal, and Regulatory (MLR) workflows within Veeva PromoMats for enterprise pharma clients (Novo Nordisk, Haleon, ALK), shortening review cycles and improving compliance adherence.
• Built Databricks and Azure Data Factory pipelines and BI/analytics models that prepare client data for downstream AI use cases.
• Implemented a data-driven approach to monitor and optimize the software development lifecycle, tracking delivery metrics across the release pipeline and reducing time to market for software releases.
• Helped productize the solution so it can be deployed as a repeatable offering across multiple pharma clients.

An industrial AI platform built around digital twins of manufacturing and industrial operations. Industrial clients generate large volumes of time-series data from their equipment and processes but lack the infrastructure to turn it into predictions and decisions. The platform ingests time-series data from each client's digital twin and runs machine learning models on top of it, providing operations teams with predictive insights. It is a seed-stage B2B product sold to industrial companies.

• Built and deployed new machine learning capabilities in the platform, supporting client acquisition valued at EUR 130K in ARR.
• Built the ML infrastructure and model-serving layer for time-series data coming from clients' digital twins.
• Developed Python and SQL pipelines on AWS and PostgreSQL to ingest and prepare time-series data for model training and inference.
• Containerized model-serving services with Docker to make deployments repeatable across client environments.

A data product for the EMEA business of a global creative and marketing software company that tracks monetization across newly defined customer cohorts. Sales and marketing teams lacked a reliable, consolidated view of how different customer segments converted and generated revenue, which made it hard to decide where to invest. The product combines customer and revenue data into cohort models and analytics that help you target campaigns and shape clients' go-to-market strategy across the region.

• Built the cohort-analysis data models end-to-end on Databricks and Spark, supporting the EMEA business in driving ~$120M in net new ARR.
• Led an end-to-end data workload behind a new monetization strategy that unlocked $1M+ in ARR per quarter.
• Redesigned the sampling data extraction process, increasing marketing campaign scope by 2.5x.
• Advised on the creation of a data framework that optimized cloud costs by 40%.
• Built and orchestrated PySpark pipelines with Airflow to keep cohort and revenue data up to date for stakeholders.

A customer-facing reporting platform for a travel startup whose B2B clients needed reliable, up-to-date reports on their travel activity and providers. Operational data was previously reloaded in full every day, which was slow and costly and caused reports to lag behind the business. The platform ingests operational data incrementally into a cloud data warehouse, validates it through layered data-quality checks, and serves reporting models directly to the startup's customers. It was designed to run on a small startup budget without heavy orchestration or paid monitoring tools.

• Designed and built the customer-facing reporting platform end-to-end as the sole data engineer, replacing daily full reloads from MongoDB (via AWS Glue) with an S3 landing zone and Snowpipe incremental loading, cutting compute costs and enabling near-real-time reporting.
• Negotiated the stack and costs with the client and vendors, choosing orchestration inside Snowflake over Airflow/Prefect to keep the platform lean.
• Designed a layered (landing / validated / reporting) data-quality framework with schema, uniqueness, volume-anomaly, and referential checks, plus reconciliation of report totals with the source system.
• Built a lightweight in-house run ledger (stage, input, output, status, error per step) for pipeline troubleshooting, avoiding the cost of paid monitoring tools such as Datadog or Grafana.
• Introduced infrastructure-as-code with Terraform, reducing time to deployment, and set documentation standards that improved knowledge transfer.

A consultant-matching solution for a data and cloud consultancy that receives a steady flow of RFPs (requests for proposal) from prospective clients. Matching each RFP to consultants with the right skills was a manual process that relied on staff expertise and slowed responses to new business. The solution uses a fine-tuned language model to extract skills, technologies, and domain entities from RFP documents and automatically recommends the best-fit consultants, which are exposed to internal teams through REST APIs.

• Researched, fine-tuned, and deployed a Hugging Face LLM for named entity recognition (NER) and domain adaptation, automating consultant recommendations for incoming RFPs and improving operational efficiency.
• Developed and deployed REST APIs that exposed the models to internal teams, bridging the gap between data science and cloud deployment.
• Built custom data transformers for the TensorFlow/Keras machine learning workflows.
• Designed and developed graphical components for the company's internal data science framework.

A B2B e-commerce and logistics platform that lets manufacturers and distributors sell, ship, and get paid online. Its finance team needed accurate visibility into cost of goods sold (COGS), accounts payable and receivable, and payment disputes, but the underlying data was fragmented and error-prone. The data platform consolidates AP/AR data, tracks COGS, and supports decisions on payments, disputes, and profitability reporting.

• Built an accounts payable (AP) data platform that improved cost tracking and data availability by 25%.
• Ran data-quality checks on AP data, tracking and fixing issues worth $100K per month.
• Automated profit and loss (P&L) reporting with Python and SQL on AWS, giving the finance team timely data for decisions on payments and disputes.

An SAP e-procurement integration program for a large Latin American telecommunications group that was absorbing newly acquired companies across the region. Each acquisition brought its own procurement processes and systems that had to be migrated to the group's source-to-pay platform without disrupting operations. The team integrated the acquired companies' procurement operations into SAP and automated repetitive back-office work with robotic process automation (RPA).

• Led an integration development that brought an acquired company's entire source-to-pay process into the group's SAP e-procurement system.
• Built RPA automation workforces and integrated them into the existing infrastructure, increasing team productivity by 20%.
• Designed and implemented SQL-based interfaces for data transfers between SAP and surrounding systems.
• Set up monitoring and reporting on the interfaces, improving data consumption and performance.
• Negotiated technical blueprints and implementation strategies with project stakeholders.