Fernando Bellido Pazos
Mathematician and Computer Scientist
I am an AI Engineer & Technical Leader with a double background in Mathematics and Computer Science (UPM, GPA 9.4/10) and an MSc in Applied & Computational Mathematics (UC3M). I bridge mathematical rigor with cloud-native engineering to design deterministic, enterprise-grade AI platforms that solve high-stakes operational challenges.
Currently at DXC Technology, I lead engineering initiatives spanning Multi-Agent orchestration (AutoGen, Google ADK, MCP), deterministic Generative AI presales pipelines (>95% cycle-time reduction), and automated SOX compliance audit systems. Trusted by enterprise clients and cross-functional teams, I specialize in translating complex business requirements into inspectable, high-throughput AI architectures with strict quality gates and zero-tolerance governance.
Beyond engineering, I cultivate relentless discipline and mental endurance through competitive athletics (CrossFit, Spartan Race, Hyrox, and open-water marathon swimming) and 4+ years as an international orchestral violinist.
Core Expertise: Multi-Agent Systems Generative AI & LLM Architecture Applied Mathematics Azure Cloud Architecture Python Enterprise Governance & MLOps
Experience
DXC Technology
1 AI Solutions Engineer - GenAI & Architecture
Led the functional analysis, architecture, and end-to-end development of an enterprise Generative AI pipeline to accelerate technical presales—replacing manual RFP/tender analysis with a deterministic, schema-first structured extraction and multi-agent auditing system (>95% time savings).
View technical details & architecture Hide technical details
- Architecture & Lifecycle: Mapped the opportunity lifecycle across 8 stages (Phases 0–7) and designed a modular suite of 9 capabilities to mitigate critical risks, including hidden penalties, tone/style inconsistencies, and bid disqualification caused by envelope-crossover errors.
- Deterministic (Schema-First) Extraction: Developed an engine utilizing long-context LLMs and layout-aware document parsing (Docling/PyMuPDF). Bypassed vector RAG in favor of strict, self-healing Pydantic v2 data contracts, structuring 450+ items with end-to-end traceability and verbatim citations.
- Multi-Agent Evaluation: Designed a virtual review committee (Neutral Arbiter alongside technical and financial risk evaluators) to audit 150+ page proposals against tender specifications, delivering clause-by-clause verdicts with detailed rationale and textual evidence.
- Governance & Human-in-the-Loop: Implemented an inspectable runtime featuring quality gates (Gate M4), enterprise cloud isolation (zero-training data policies), and end-to-end audit logging.
- Impact & Empirical Validation: Validated the MVP on live procurement bids (34-page RFP vs. 153-page proposal), slashing compliance matrix generation from 16–47 hours to under 10 minutes (>95% time savings) with 100% coverage across 320+ requirements (under 7% variance).
2 AI Auditing System
Technical Delivery Lead for an agentic SOX compliance auditing solution built with Google ADK, providing automated evidence validation for Control Owners and significantly reducing auditor overhead.
View technical details & architecture Hide technical details
- Technical Leadership: Led the technical delivery, managing developers and coordinating with project management while engineering the core AI solution.
- Agentic Evidence Analysis: Built the agentic system using Google ADK to analyze uploaded audit evidences (various file formats and sizes) and determine sufficiency against SOX controls.
- Pre-Submission Validation: Generated detailed, AI-analyzed reports for Control Owners, providing pre-submission validation that was previously unavailable.
- Cost & Effort Reduction: Significantly reduced auditor time for evidence checking, resulting in substantial cost savings by minimizing required audit resources.
- Compliance Quality: Empowered Control Owners with a precise inventory and validation check, streamlining evidence preparation and improving compliance quality.
3 Enterprise AI Platform Leadership
Directed cross-functional teams to architect and deploy an enterprise Microsoft AutoGen multi-agent AI platform on Azure, managing microservices and cloud-native event-driven pipelines processing 1M+ records.
View technical details & architecture Hide technical details
- Multi-Agent Architecture: Led cross-functional teams to design and launch a Microsoft AutoGen-powered multi-agent AI platform, managing microservices and cloud-native architecture on Azure.
- Intelligent Agents: Architected intelligent agents (geospatial, claims, decision automation) at enterprise scale using Azure OpenAI GPT-4o.
- Event-Driven Pipelines: Designed event-driven data pipelines (Azure EventHub, 99.9% reliability), processing over 1M records.
- Roadmap & Deployment: Owned end-to-end delivery from technical roadmap to enterprise production deployment.
- DevOps & Model Validation: Built DevOps pipelines, CI/CD automation, and AI model validation frameworks for scalable production systems.
- Team Mentorship: Mentored teams on advanced AI, multi-agent orchestration, and secure execution environments.
4 Agentic Traffic Management Automation Platform
Architected and delivered a multi-agent geospatial planning engine automating temporary traffic management layouts from high-level roadwork requests using Google ADK, Agent2Agent, MCP, PostgreSQL, and Kubernetes.
View technical details & architecture Hide technical details
- Geospatial Planning Engine: Designed and implemented a multi-agent geospatial planning engine to automate temporary traffic management layouts from high-level roadwork requests using Google ADK, Agent2Agent, MCP, PostgreSQL, Azure, and Kubernetes.
- Orchestration Layer: Built an orchestration layer coordinating specialised agents and shared MCP tools, ensuring deterministic, auditable workflows and clean separation between business logic, standards, and geospatial computation.
- Enterprise Integration: Integrated the platform with existing enterprise systems (including CRM and work management tools) via dedicated adapters, enabling end-to-end automation from request intake to deployment-ready traffic plans.
- Standardization & Quality: Improved delivery speed and consistency of traffic management designs by standardising computations, reducing manual configuration effort, and lowering the risk of human error in layout definition.
IDOM | Consulting, Engineering, Architecture
1 System Innovator
Drove research and development across Digital Twins (Autodesk Tandem & Azure IoT), internal RAG systems, BERT NLP classification, and dockerized AI development infrastructure.
View technical details & architecture Hide technical details
- Digital Twin PoC: Developed a Proof-of-Concept Digital Twin utilizing Autodesk Tandem and Azure IoT Hub, in conjunction with Azure Stream Analytics and Azure ML Studio, to showcase advanced capabilities outperforming traditional human monitoring with real-time data integration and predictive analytics.
- Internal RAG & AI Companion: Directed theoretical R&D of an innovative internal Retrieval Augmented Generation (RAG) system, significantly enhancing organizational language understanding and creating an internal AI-driven companion.
- BERT NLP Classification: Pioneered the development of an AI classifier using NLP and Google’s BERT model, complemented by an ontology-based approach for sophisticated categorization from comprehensive project descriptions.
- Automated CV Generator: Innovated a CV generator grounded in S&P’s Success Factors system, featuring data ingestion, AI-driven paragraph generation, template compilation, and polished output.
- AI Infrastructure: Architected and deployed a cutting-edge AI development infrastructure, encapsulating a fully dockerized framework with real-time resource monitoring, Traefik routing, and advanced AI tools like Stable Diffusion.
2 Consulting – Digital Transformation
Engineered automated SAP tabular data pipelines with Azure Document Intelligence and Phi-3.5, predictive asset maintenance models, and regional statistical property valuation systems.
View technical details & architecture Hide technical details
- SAP Data Pipeline: Developed an AI-automated pipeline for processing tabular data into specific SAP naming conventions using Azure Document Intelligence and Azure Phi-3.5 cloud deployment, significantly reducing data processing time and boosting data integration consistency.
- Intelligent Asset Maintenance: Spearheaded the development of AI and mathematical models, revolutionizing asset maintenance from reactive to proactive using S&P’s Intelligent Asset Management and IoT modules.
- Cadastre Statistical Valuation: Played a pivotal role as a Statistician developing the cadastre system with the registry office, innovating a unique calculation method for assessing house values across the Basque Country.
Ontology Engineering Group
1 Data Analyst and Developer Intern
Developed knowledge graph construction pipelines from institutional academic repositories using RDF and SPARQL, automating metadata extraction to increase ingestion efficiency by 40%.
View technical details & architecture Hide technical details
- Knowledge Graph Construction: Developed a system to construct knowledge graphs from academic papers in institutional repositories, enhancing advanced querying capabilities for Open Science principles.
- Metadata & Relationship Extraction: Analyzed large datasets of academic publications to extract meaningful metadata and relationships, improving data accessibility and facilitating deeper research insights.
- Ontology Modeling: Implemented ontology-based data models to represent complex scholarly information, enabling better interoperability and integration across data sources.
- Automated Ingestion: Optimized data ingestion processes by automating metadata extraction and semantic annotation, increasing efficiency by 40% compared to manual methods.
- Semantic Web Stack: Utilized RDF, SPARQL, and Python semantic libraries to enhance the group’s semantic web capabilities.
Client Recommendations & Endorsements
“Fernando has demonstrated a professional caliber that goes well beyond technical execution. From day one, he stood out for his proactivity and ability to propose ideas that delivered real value, anticipating needs before they were even raised. One of the most valuable aspects of his contribution was taking over a product that arrived in a very deficient state—he took on the challenge with determination and sound judgment, turning it into a fully operational and reliable solution that earned the direct trust of the project sponsor. In meetings, his ability to moderate, structure debate, and redirect complex discussions was a fundamental asset. I recommend him without reservation for any role requiring not only technical competence, but also leadership and vision.”
- Scope & Project: Leadership and core architecture across the autoSOX initiative.
- Endorsement Highlights: Exceptional stakeholder management, end-to-end product turnaround, and executive leadership.
“I would like to highlight in particular his technical and functional proficiency, translating demanding internal control requirements into stable, well-documented solutions aligned with audit criteria. His availability and willingness to help—especially during peak periods of the SOX audit schedule—always responded with dependability and commitment. His contribution had a real impact on the project’s outcome and proved he can take on greater responsibilities than initially planned.”
- Scope & Project: Auto SOX (Automated SOX control testing and evidence validation).
- Endorsement Highlights: Technical solvency, proactive functional enhancements, and high reliability under peak audit cycles.
María Vega Laguna Díaz
Data Scientist (Digital Hub) · Ferrovial
June 2026
View Recommendation Letter (PDF)
“Fernando is decisive and manages priorities effectively. He has made a significant qualitative leap across initiatives like Contract Management, Lane Closure, and AutoSOX: evolving from a purely technical engineer into the key point of reference for the business team. He possesses a remarkable ability to listen to business needs, understand them, and successfully translate them into technical implementation—bringing tremendous value. His work goes well beyond the standard functions of an AI engineer, demonstrating the capabilities needed to maintain an end-to-end project vision.”
- Scope & Initiatives: AutoSOX, Lane Closure, and Contract Management (Ferrovial Digital Hub).
- Endorsement Highlights: Business-to-technical translation, cross-project agility, and functional leadership.
Certifications
Claude Partner Badge - Claude Code
Anthropic
Issued Aug. 2026 · Expires Feb. 2027
- Issuer: Anthropic
- Credential ID:
143ab35e-21c3-4988-bfee-b0150a67f5f
Claude Certified Associate - Foundations
Anthropic
Issued Aug. 2026 · Expires Aug. 2027
- Issuer: Anthropic
- Credential ID:
2ebf03ef-f068-4a42-b138-07f6319d1295
Microsoft Certified: Azure AI Fundamentals
Microsoft
Issued Jan. 2025
- Issuer: Microsoft
- Credential ID:
E7243975041289D2
Education
Carlos III University of Madrid
Madrid, Spain
Sept. 2022 - June 2023
uc3m.es/master/applied-mathematics
MSc in Applied & Computational Mathematics [English]
-
GPA: 8.38/10 with 1 Honourable Mention in Advanced Analysis.
-
Relevant Coursework: Biological Inspired Artificial Intelligence, Modelling and Nonlinear Analysis, Optimization, Stochastic Equations for Finance and Biology, Complex Biological and Socioeconomic systems, Computational Linear Algebra, Computational techniques for Differential equations, Methods for Nonlinear differential equations, Applied discrete mathematics.
Bachelor’s degree in mathematics & computer science [Spanish]
-
GPA: 9.4/10 with 14 Honourable mentions.
-
Relevant Coursework: Data Analysis, Software Engineering, Operating Systems, Artificial Intelligence, Databases, Modelling, Programming (Imperative, Object Oriented, Functional, Logic, Systems), IT Project Management, Topology, Calculus, Algebra, Pure Mathematics.
Projects
Basic Numerical Methods in Python - Student Oriented (June 2023)
My end of master’s thesis aims to develop an open-source PyPi project that includes a scripted version of the numerical methods taught in universities and a Jupyter-integrated GUI for real-time visualization of those. The goal is to help students understand the methods and promote learning Python instead of MATLAB, which is not as popular or accessible as Python in the current job market.
-
This project was awarded the ‘Most Innovative Project of 2023’ by Logicalis in the Logicalis Innovation Challenge, recognizing its potential to significantly impact educational practices and student learning experiences.
- Tech: Python, Jupyter, Numpy, BQPlot, iPyWidgets
- Link: github.com/fbpazos/Trabajo-Fin-Master
Sturm Liouville Problem and Models based on differential equations on the field of Acoustics (June 2022)
Pursued a comprehensive study of the Sturm-Liouville theory and modelled stringed musical instruments to explore the relationship between mathematics and music. The aim of the research was to answer questions such as: what factors contribute to the distinct sound of instruments like a violin and a piano, and how does mathematics play a role in the creation of music.
- Tech: Python, Latex, Matplotlib, Plotly, Sympy, NumPy, Jupyter, Differential Equations, Functional Analysis.
- Link: oa.upm.es/70879
Honors and Awards
- Logicalis Innovation Challenge (2023)
- Cambridge English Advanced Certificate (C1)
- Academic Excellence Grant (2021 & 2022)
- Academic Excellence Award for Representative Agents of Banco Mediolanum (2022)
A Little More About Me
Alongside my passion for mathematics and artificial intelligence, I channel deep discipline, creativity, and relentless drive through athletics and the arts:
CrossFit & High-Performance Fitness
A dedicated CrossFit athlete focused on functional strength, metabolic conditioning, and peak physical resilience. I thrive in high-intensity environments that push mental and cardiovascular limits—treating physical discipline as the foundation for cognitive focus and engineering problem-solving.
- Competitive Races & Fitness Challenges:
- Spartan Race — Obstacle Course Racing (OCR) demanding total endurance, agility, and grit.
- Hyrox Bilbao — Global functional fitness competition combining 8 km running with 8 functional workout stations.
- Deka Fit Madrid & Deka Fit Zaragoza — Decathlon of functional fitness testing multi-domain athletic capacity.
Violin & Classical Music
With over 4 years of dedicated study and 2 hours of daily practice, I have cultivated advanced violin technique, orchestral performance experience, and music theory rigor.
As an orchestral violinist, I had the honor of performing internationally at the European Youth Music Week (EYMW) 2022 in Oakham (United Kingdom) and EYMW 2023 in Sondershausen (Germany).
- Featured Performances & Concert Recordings:
- Partita II (BWV 1004) Allemande — J.S. Bach (June 2022): Watch on YouTube
- Sonata for Violin & Piano (KV. 304) — W.A. Mozart (June 2022): Watch on YouTube
- Salut d’Amour (to Carice) — Edward Elgar (December 2022): Watch on YouTube
- Double Violin Concerto (BWV 1043) — J.S. Bach (June 2022): Watch on YouTube
- Opus 34, First Movement — Oskar Rieding (December 2021): Watch on YouTube
Swimming & Philanthropy
Swimming has been an essential pillar of my lifestyle for over 5 years, fostering aerobic capacity, mental clarity, and athletic endurance.
- Open Water & Community Impact: Actively training for a 5 km open sea swimming semi-marathon in support of ALS (Amyotrophic Lateral Sclerosis) research and patient advocacy—uniting endurance sports with meaningful social impact.