Bilbao, Spain 🇪🇸 | Data Science, Software Engineering & Enterprise Data Management
I operate at the intersection of Data Science, Software Engineering, and Enterprise Data Management.
In 2026, enterprise software and AI systems don't fail from lack of raw model capacity — they fail due to untamed data quality, brittle architecture, and lack of deterministic engineering. I combine rigorous Python software engineering with DAMA-DMBOK data management principles to transform complex data into resilient, scalable, production-grade applications.
- Software & Systems Engineering: Architecting scalable microservices, low-latency REST and WebSocket streaming APIs with FastAPI, runtime data contracts with Pydantic v2, and strict test-driven development (
pytest). - Data Science & Modeling: Exploratory data analysis (EDA), statistical inference, predictive modeling, and feature engineering across structured and unstructured industrial datasets (Pandas, NumPy, Scikit-Learn, PyTorch).
- Open Source Foundations: Active open-source collaborator contributing targeted core features, multi-chunk streaming engines, and comprehensive localization across foundational AI repositories (LiteLLM, Hugging Face).
- Enterprise Data Management (CDMP®): Grounded in the DAMA-DMBOK framework — enforcing Data Quality dimensions, Data Governance, Metadata Management, and Master Data Architecture (MDM).
- CDMP® — Certified Data Management Professional (DAMA International):
The global gold-standard certification in enterprise data management and the DAMA-DMBOK (Data Management Body of Knowledge). Specializing in Data Quality metrics, Data Governance frameworks, and Enterprise Data Architecture.
| Domain | Technologies & Frameworks |
|---|---|
| Software & Systems |       |
| Data Science & ML |        |
| Data Governance |     |
| DevOps & Tooling |     |
-
- PR #2953: Made
space_sdkconfigurable inCodeAgent.push_to_hub(), resolving deployment errors (402 Payment Required) on static spaces and custom hub configurations. - PR #2952: Documented missing execution parameters (
authorized_importsandmax_print_outputs_length) inlocal_python_executor.evaluate_python_codewith regression test coverage.
- PR #2953: Made
-
- PR #756: Synchronized Unit 2.1 (
smolagentsand code agents) with upstream reference curriculum and fixed navigation indices.
- PR #756: Synchronized Unit 2.1 (
-
- Lead contributor to the Spanish curriculum with 29 dedicated pull requests covering 100% of the entire course: Transformers, tokenization, distributed inference (vLLM PagedAttention, TGI, llama.cpp), Argilla data curation, Gradio demos, and Reasoning Models (DeepSeek R1 / GRPO in TRL).
Expand list of 29 Hugging Face Course Pull Requests
Module Pull Requests Topic Chapter 1 #1330 Certification Exam ( chapter1/11.mdx)Chapter 2 #1325, #1329, #1333, #1349 Pipelines, Models, TGI, vLLM, llama.cpp & Quiz Chapter 3 #1330 Certification Quiz ( chapter3/7.mdx)Chapter 4 #1331, #1332, #1334, #1335, #1336 Model Hub, Pretrained Models, Model Cards & Quiz Chapter 7 #1350, #1351, #1352, #1353 Token Classification, Masked LM, Translation, Summarization, CLM, QA & LLMs Chapter 8 #1337, #1338, #1339, #1341 Forums, PyTorch & TensorFlow Training Pipeline Debugging & Issues Chapter 9 #1342, #1343, #1344, #1345 Gradio Interfaces, Sharing Demos, Spaces Integration, Blocks & Quiz Chapter 10 #1346, #1347, #1348 Argilla Dataset Annotation, Setup, Verification & Quiz Chapter 11 #1354 Fine-tuning Large Language Models (Sections 1 to 7) Chapter 12 #1355, #1356 Reasoning Models, DeepSeek R1 & GRPO in TRL & Unsloth Events #1340 Community Launch Events & Challenges
- BerriAI/litellm:
- PR #44439: Fixed tool-call streaming fragmentation in
ChunkProcessor.get_combined_tool_content, preventing truncated function names and IDs during multi-vendor proxy streaming. Added regression unit test coverage. - PR #44459: Set status description on OpenTelemetry error spans (#44184), capturing root-cause exception messages in observability collectors.
- PR #44439: Fixed tool-call streaming fragmentation in
-
tidyverse/dplyr (PR #7856):
Addressed core semantic documentation (#6968) forfilter(), clarifying simultaneous vs. sequential evaluation when conditions compute summary and aggregation metrics. -
punkpeye/awesome-mcp-servers (PR #15720):
Contributed the Enterprise Data Governance MCP server implementing DAMA-DMBOK quality dimensions and metadata tools for Claude/Desktop AI agents.
A curated index of canonical, high-impact open-source repositories shaping metallurgical vision, scrap characterization, industrial thermography, enterprise data governance, and high-performance analytical systems:
Canonical Ecosystem Repository:
For the comprehensive, deep directory with architectural diagrams, combustion solvers, and benchmark datasets, explore:
ioseba/awesome-industrial-metallurgy-ai — Awesome list covering Secondary Aluminum Recycling, LWIR Radiometric Thermography, Refractory Computer Vision, and Casthouse IIoT.
Metallurgy, Scrap Metal Characterization & Steel/Aluminum Inspection
| Repository | Discipline | Highlights |
|---|---|---|
| ioseba/awesome-industrial-metallurgy-ai | Flagship Index Awesome List |
Canonical curated catalog of repositories, tools, datasets, and standards for industrial metallurgy, aluminum recycling, thermography, and refractory AI. |
| zichengzichengzi/Scrap_Steel_Image_Dataset | Scrap Grading Object Detection |
Specialized open dataset and annotations for automated scrap steel grading, sorting, and hazardous non-metallic impurity detection. |
| stepanje/MPDD | Metal Parts AD Manufacturing Flaws |
Metal Parts Defect Detection benchmark dataset for computer vision, covering painted, non-painted, and machined metal surfaces with irregular defects. |
| khornlund/severstal-steel-defect-detection | Defect Segmentation Hot-Rolled Steel |
Canonical benchmark implementation for the Severstal steel defect challenge. PyTorch segmentation of pitting, inclusion, and surface roll cracks in continuous manufacturing lines. |
| YeahHuang/Al_surface_defect_detection | Aluminum Defect AI Extrusion & Sheet |
Benchmark vision pipelines for industrial aluminum surface defect classification and segmentation (coating spots, non-metallic inclusions, peeling, and scratches). |
| siddhartamukherjee/NEU-DET-Steel-Surface-Defect-Detection | NEU-DET Benchmark Surface Flaws |
Comprehensive deep learning implementation on Northeastern University's steel surface defect dataset (crazing, inclusion, patches, pitted surface, rolled-in scale, and scratches). |
| spectralpython/spectral | Hyperspectral AI Alloy Classification |
Fundamental Python package for hyperspectral and multispectral image processing, used in automated scrap sorting lines to distinguish non-ferrous aluminum alloys (series 5xxx vs 6xxx). |
| xraypy/xraylarch | XRF / LIBS Spectroscopy Metal Characterization |
Open-source Python library for X-ray fluorescence (XRF) and spectroscopy data analysis, essential for scrap elemental composition and impurity verification. |
| DLR-WR/libskit | LIBS Spectroscopy Alloy Chemistry |
Software toolkit for Laser-Induced Breakdown Spectroscopy (LIBS) spectral processing and fast elemental quantification in inline metal scrap analysis. |
Aluminum Recycling Furnaces, Thermodynamics & Combustion Engineering
| Repository | Discipline | Highlights |
|---|---|---|
| Cantera/cantera | Combustion Kinetics Burner Modeling |
The premier open-source chemical kinetics, thermodynamics, and transport suite. Used for oxy-fuel / air-fuel burner modeling, flame temperature calculations, and NOx emission modeling in melting furnaces. |
| pycalphad/pycalphad | CALPHAD Equilibrium Slag & Dross Chemistry |
Computational thermodynamics for multicomponent alloy phase diagrams. Essential for simulating aluminum-salt flux (NaCl/KCl) interactions, dross formation, and scrap melting thermodynamics. |
| OpenFOAM/OpenFOAM-dev | Computational Fluid Dynamics Combustion & Melt Flow |
The open-source CFD benchmark for simulating turbulent burner combustion, furnace flue gas recirculation, and molten aluminum fluid dynamics. |
| lanl/truchas | Multi-Physics Simulation Metal Casting & Solidification |
Developed by Los Alamos National Laboratory: 3D multi-physics simulation suite specialized in metal casting, fluid flow, heat transfer, and phase-change solidification. |
| CalebBell/ht | Heat Transfer Radiation & Convection |
Comprehensive heat transfer calculation engine. Crucial for refractory wall thermal resistance, shell heat dissipation, and radiation exchange between burner flames and molten aluminum baths. |
| CalebBell/thermo | Chemical Thermodynamics Process Simulation |
High-fidelity pure component and mixture thermodynamic properties library (vapor-liquid-solid equilibrium, enthalpy, and thermal conductivities). |
| oemof/tespy | Thermal Plant Balances Waste Heat Recovery |
Simulation toolkit for mass and energy balances in thermal engineering systems. Ideal for modeling furnace off-gas lines, recuperators, and heat recovery loops in aluminum recycling plants. |
| CoolProp/CoolProp | Thermophysical Properties Flue Gas Fluids |
High-accuracy fluid and flue gas thermodynamic property formulations across extreme temperature and pressure regimes. |
Industrial Computer Vision, Surface Defect Detection & Anomaly Inspection
| Repository | Discipline | Highlights |
|---|---|---|
| open-edge-platform/anomalib | Unsupervised AD PatchCore / PaDiM |
Intel's flagship industrial anomaly detection library. Benchmark algorithms for spotting visual manufacturing defects without requiring vast labeled defect datasets. |
| Charmve/Surface-Defect-Detection | Defect Datasets Metal Inspection |
Canonical global catalog of surface defect detection datasets, deep learning architectures, and papers for steel, aluminum, and ceramic substrates. |
| ultralytics/ultralytics | Real-Time Detection YOLO11 |
The industry standard for high-throughput real-time object detection and instance segmentation on factory floors and scrap sorting lines. |
| facebookresearch/segment-anything | Foundation Models Zero-Shot SAM |
State-of-the-art segmentation foundation model for precise boundary delineation of refractory spalling, cracks, and structural erosion. |
| roboflow/supervision | Vision Utilities Tracking & Polygon |
Reusable computer vision utilities for inference annotation, polygon zone filtering, and object tracking across production camera feeds. |
Thermal Imaging, Radiometric Vision & High-Temperature Monitoring
| Repository | Discipline | Highlights |
|---|---|---|
| LJMUAstroecology/flirpy |
Radiometric IR FLIR Cores
|
Python library for programmatic control of FLIR cameras and extraction of raw radiometric temperature arrays ( |
| SanNianYiSi/thermal_parser |
Radiometry IR Metadata
|
Lightweight parser for radiometric temperature matrix reconstruction and emissivity correction from raw thermal images. |
| OpenThermal/libseek-thermal |
Thermal SDK LWIR Sensors
|
Open-source driver and SDK for Seek Thermal compact long-wave infrared (LWIR) camera cores. |
| ESA-PhiLab/PyRawS |
Thermal Anomalies Hotspot Detection
|
Algorithmic toolkit for high-temperature surface thermal anomaly detection, active wildfire/hotspot radiance tracking, and saturation handling. |
Industrial IoT, OPC-UA & Telemetry Sensor Fusion
| Repository | Discipline | Highlights |
|---|---|---|
| FreeOpcUa/opcua-asyncio | OPC-UA AsyncIO Telemetry |
Modern asynchronous OPC-UA stack in Python for bi-directional communication with industrial PLCs, furnaces, and plant SCADA systems. |
| frangoteam/FUXA | Web SCADA HMI / Dashboard |
Open-source web-based process visualization and SCADA/HMI software supporting OPC-UA, Modbus, MQTT, and Siemens S7. |
| Serial-Studio/Serial-Studio | Telemetry Viz Multi-Sensor Bus |
High-performance multi-platform real-time telemetry dashboard for serial, CAN bus, Modbus, and MQTT industrial data streams. |
| taosdata/TDengine | Industrial Time-Series IIoT Database |
Purpose-built, high-throughput time-series database engineered for high-frequency sensor readings, pyrometers, and furnace telemetry. |
Enterprise Data Governance & Data Quality (CDMP® Alignment)
| Repository | Discipline | Highlights |
|---|---|---|
| OpenMetadata | Metadata Lineage Discovery |
Open Context Layer offering end-to-end data lineage, active data governance, automated data contracts, and unified team collaboration. |
| DataHub | Data Catalog Observability |
LinkedIn-originated stream-based metadata platform designed for high-scale enterprise search, data mesh topologies, and compliance tracking. |
| Great Expectations | Data Quality Automated Testing |
Industry benchmark for declarative data validation, automated pipeline assertions, and living data documentation (Data Docs). |
| OpenLineage | Data Lineage Open Standard |
Standardized framework for collecting operational runtime lineage metadata across Airflow, Spark, dbt, and modern orchestrators. |
| Soda Core | Data Contracts SodaCL |
Developer-friendly declarative data contracts engine and quality monitor integrated directly into transformation pipelines. |
High-Performance Data Engines, Analytics & Visualization
| Repository | Discipline | Highlights |
|---|---|---|
| DuckDB | Embedded OLAP Vectorized SQL |
Fast, in-process analytical SQL database engine with zero external dependencies; the canonical "SQLite for Analytics" for fast local & object storage querying. |
| Polars | Rust DataFrames Out-of-Core |
Blazingly fast multi-threaded columnar DataFrame library built in Rust, leveraging memory locality and lazy query evaluation. |
| Apache Arrow | Columnar In-Memory Zero-Copy |
The global cross-language substrate powering high-throughput in-memory data processing, interoperability, and analytical compute. |
| dbt-core | Analytics Engineering Data Modeling |
The foundational standard bringing software engineering rigor (git, testing, CI/CD, documentation) to SQL-based data transformations. |
| Apache Superset | Enterprise BI Exploratory Viz |
Highly scalable, modern enterprise business intelligence and data visualization platform supporting any SQL dialect. |
Modern AI Systems & Edge Inference Engines
| Repository | Discipline | Highlights |
|---|---|---|
| openvinotoolkit/openvino | Edge AI Intel Inference |
Intel's toolkit for optimizing and deploying deep learning models with ultra-low latency on industrial edge hardware and IPCs. |
| microsoft/onnxruntime | Cross-Platform ML Hardware Acceleration |
Universal runtime for accelerating PyTorch and ONNX models on CPU, CUDA, TensorRT, and edge NPUs. |
| vLLM | LLM Inference PagedAttention |
High-throughput, memory-optimized serving engine with continuous batching for production self-hosted LLM deployments. |
| LiteLLM | AI Gateway Proxy & Governance |
Enterprise AI gateway unifying 100+ LLM backends behind an OpenAI-compatible interface with load balancing, rate limiting, and cost controls. |
| Transformers | Foundation Models PyTorch / JAX |
The definitive library for downloading, training, fine-tuning, and evaluating state-of-the-art vision, audio, and language models. |
