Back to home

Carlos David Romero Núñez

Tech Lead · Data & AI Platform Architect · Senior Data Engineer

Data & AI platform architect, self-taught. Twelve years writing code: I started fullstack, specialized in data —ETL, data lake, data warehouse, lakehouse, Medallion architecture, dimensional modeling—, then in the infrastructure behind it: AWS, Azure, Terraform, on-premises. I work in Python, SQL and Go. Today I specialize in agentic development: agents on frontier models, and local AI with quantized models and MoE —mixture-of-experts— architectures.

Experience

  1. Stack Overflight

    Tech Lead / Data & AI Platform Architect

    Sep 2025 – present (1 yr)

    I lead the data architecture, infrastructure and product of a European airline's BI platform —a confidential client—, plus the consultancy's own products. I designed the five frameworks the operation runs on —tf-admin for infrastructure, Core Bronze, Silver and Gold for the Medallion pipeline, bi-solution for BI—, alongside the 13 AI agents that operate them daily, with 83 skills between them: that automation lets the team sustain the platform on 60% less headcount than projected. The airline's BI went from zero to the company's single source of data: 36 dashboards over 136 sources. The whole infrastructure runs at $4,767 a year on AWS, against an equivalent Redshift 2 to 7 times more expensive at list prices. Beyond delivery, I train the team in applied AI and data modeling fundamentals.

    Technologies: Amazon Web Services, Terraform, Medallion Architecture, Apache Superset, Grafana, PostgreSQL, DuckDB, Go (Golang), Python, Angular, cortex-ai (agentic engine), RAG, Docker

  2. Inversiones de Seguridad LTDA

    Infrastructure Consultant

    Sep 2010 – present (16 yrs) · recurring consulting, in parallel

    I moved the company from an on-premises server to Microsoft 365 and have run the tenant ever since —12 users today. Their payroll software runs on MS-DOS and couldn't be replaced, because operations depend on it: I adapted it to run virtualized on modern systems, with cloud backups and access from other sites through SharePoint Online. As internal auditor, I wrote the IT documentation for BASC 5.0.3 certification.

    Technologies: Microsoft 365 / Office 365, Next.js, Power Apps, SQL Server, Power BI, ETL / Data Pipelines, BASC

  3. Marped Group S.A.S.

    Data Engineer

    Oct 2024 – Apr 2025 (6 mos)

    I built the corporate Data Lake on the company's existing servers to replace a reporting flow that relied on hand-exported CSVs and Excel 2013 macros. I wrote the ETL in Python over reusable modules and a declarative JSON registry that defines which processes run and in what order; data lands in Parquet partitioned by year and month, so each run only rewrites recent periods. I also deployed a multi-user JupyterHub server on Ubuntu, with the Data Lake mounted and per-user and per-department isolation, so teams could run their notebooks without installing anything. Purchase planning went from ~2 hours to 24 seconds. I chose a Data Lake over a Data Warehouse for implementation speed, leaving the structure ready to migrate.

    Technologies: Data Lake, Power BI, Microsoft 365 / Office 365, SQL Server, ETL / Data Pipelines, Linux / Ubuntu Server

  4. BioGerencia S.A.S.

    Data Analysis & Reporting Advisor

    Jun 2023 – Oct 2023 (4 mos)

    I designed the entire project's data system: a fuzzy-matching algorithm —TF-IDF and cosine similarity— to reconcile two databases of educational institutions with no common identifier, and a code-built evaluation form (Google Apps Script, 90+ validated fields) to track teacher workshops. I cleaned and modeled the data in Power Query and Power Pivot, and automated with JavaScript the generation of the final reports for 135 educational institutions in InDesign.

    Technologies: GPT / OpenAI, Power Pivot, Power Query, Adobe InDesign

  5. VM Innova S.A.S.

    Executive Director / General Manager

    Oct 2018 – Jun 2023 (4 yrs 8 mos)

    I ran an office-furniture manufacturer and importer focused on public procurement —import logistics, currency trading with the dealing desk, and the full operation— and based management on automation and data from day one. I built real-time inventory control: dispatch logs cross-checked daily against a rotating section-by-section count I wrote in Python, able to audit the full inventory every day without halting operations. I cross-referenced that data with accounting and the DIAN official exchange rate to keep inventory costing and sale prices aligned in near real time, and separately built the manufacturing materials costing chain, from input to finished product. For tenders I built the full decision chain: a fast costing estimator with deviation bounded to ~9% to decide what to compete for before investing time in a full proposal, an automation that cross-checked against the RUP (Single Bidder Registry) to pre-assess legal viability, and a predictor of the optimal bid value for the electronic auction —how much to bid without exceeding the target value. I was starting AI-based analysis when I left the company. I also automated receivables, payments and the whole administrative dispatch cycle.

    Technologies: Power BI, Python, Power Apps, Microsoft Excel

  6. Aromatic Grower S.A.S.

    Product Development & Market Research Consultant

    Mar 2017 – Apr 2018 (1 yr 1 mo)

    I wrote the scraping that fed the market analysis for fresh-herb exports, and built the reports with Power Query and Power Pivot.

    Technologies: Python, Power Pivot, Power Query, Adobe Illustrator, Adobe InDesign

Training in progress

AWS Certified Solutions Architect – Associate (SAA-C03)