Daniel Cristian Fat
AI Architect designing systems that matter
I design and build enterprise AI systems: the platforms that serve models, the gateways that govern access, the agent frameworks that orchestrate autonomous work, and the governance structures that keep it all accountable. Day-to-day, that means drawing architecture diagrams, reviewing vendor proposals, aligning systems with EU AI Act requirements, and writing the technical specifications that turn strategy into working software.
My path here started in Romania, where I studied Computer Science and first encountered machine learning—building stock forecasting systems and learning the fundamentals of AI. I moved to the UK for my Master's at Lincoln, where I researched image captioning: teaching machines to understand visual content and generate natural language descriptions. That work—combining computer vision with language generation—now forms the foundation of today's multi-modal LLMs like GPT-4V and Claude's vision capabilities. After a stint as a backend developer at a startup, I joined Beazley and found my calling in the intersection of data science and enterprise systems. Over the years, I've progressed from building individual models to designing the platforms and governance frameworks that make AI work at scale.
A decade of working with AI systems has taught me that the hardest problems aren't the algorithms—they're the systems around them. Today, I focus on making AI actually work: in production, within regulatory constraints, and in ways that set engineering teams up to deliver.
What I'm Working On
Here's where my time goes. These are the problems I'm solving right now—real work, not aspirational bullet points.
AI Systems Architecture
Designing the infrastructure that makes AI usable: API gateways that route and govern LLM traffic, RAG pipelines that retrieve the right context, and agent frameworks that let autonomous systems do real work. I draw the diagrams, write the ADRs, and make sure the pieces fit together.
Governance & Compliance
Building the guardrails that let AI ship safely: EU AI Act risk assessments, model validation processes, audit trails for regulators, and reviews of third-party AI vendors. Governance isn't bureaucracy—it's how you earn the right to deploy.
Agentic AI & Orchestration
Figuring out how agents work in practice: which tasks to automate, how to coordinate multi-agent workflows, where to put human checkpoints, and how to make autonomous systems predictable enough to trust. Still early days, but this is where AI is heading.
Engineering Enablement
Making AI engineers more effective: building shared components they can reuse, documenting patterns so they don't reinvent solutions, and removing the friction that slows down delivery. Architecture is a service to the teams doing the work.
The Journey
From studying AI in Romania to architecting enterprise systems in the UK—a progression built on shipping real systems and solving increasingly complex problems.
AI Architect
Beazley Group
Designing the technical architecture for enterprise AI adoption. I own the blueprints for our GenAI platform—the API gateways that route and govern LLM traffic, the RAG infrastructure that gives models access to institutional knowledge, and the agent frameworks that enable autonomous workflows.
Beyond building, I spend significant time on governance: ensuring our AI systems align with EU AI Act requirements, conducting risk assessments on third-party AI vendors, and creating the documentation that regulators and auditors need. A key part of my role is enablement—building reusable components and patterns so AI engineers can move fast without reinventing solutions.
Senior Data Scientist
Beazley Group · London
Stepped into technical leadership, managing a team of data scientists and engineers while staying hands-on with the hardest problems. Built and deployed an enterprise sanctions screening system using NLP and knowledge graphs—a project that fundamentally changed how the compliance team operated.
Owned the full MLOps lifecycle: model development, validation, deployment, monitoring, and governance. Worked closely with actuaries and underwriters on pricing models, learning the domain deeply. This is where I started thinking like an architect—seeing beyond individual models to the systems that connect them.
Data Scientist
Beazley Group · London
My introduction to insurance and specialty risk. Built machine learning models for behavioural risk pricing, developed entity resolution systems to match records across fragmented data sources, and created NER pipelines to extract structured information from unstructured policy documents.
Learned to work in a heavily regulated environment where model decisions have real financial consequences. Started building APIs and thinking about how models integrate into larger systems—the first steps toward an architectural mindset.
Software Developer
Accommodation.co.uk · Lincoln
Startup environment, small team, big ambitions. Joined to help rebuild the backend infrastructure for a student accommodation platform. Worked across the full stack but focused primarily on Python and Django—building RESTful APIs, optimising database queries, and integrating with third-party services.
The pace was intense and resources limited, which taught me to ship fast, make pragmatic trade-offs, and understand that software exists to solve real problems for real users. This experience grounded my later work—even as systems got more complex, I never forgot that the goal is to deliver value.
Education
The academic foundation—where I first discovered machine learning, built my first AI systems, and learned to think computationally.
MSc Computer Science
University of Lincoln
Research Project: Image Captioning
Developed an efficient approach to image captioning where machines generate human-like descriptions, accounting for relationships between elements in the image.
Machine Learning
Theoretical fundamentals and practical application of supervised, unsupervised, reinforcement and evolutionary learning—handwritten digit recognition, machine translation.
Mobile & IoT
Built a Smart Attendance System using Raspberry Pi, NFC sensors, and Xamarin mobile app with REST API.
BSc Computer Science
University of Alba Iulia
Research Project: Stock Market Forecasting
Built a platform for analysing and comparing company stock data with short-term price prediction capabilities.
Artificial Intelligence
Decision-making, problem-solving and learning abilities in software agents—theoretical fundamentals and practical applications.
Machine Learning & Pattern Recognition
Foundational ML algorithms and pattern recognition techniques.
Computer Technician
Technical College Alexandru Domșa
Electrical, Electronic and Communications Engineering Technology. Where it all started—hardware fundamentals, basic programming, and the first spark of interest in computing.
Technical Toolkit
A pragmatic stack built from years of shipping real systems—always choosing the right tool for the job.
Languages
GenAI & LLMs
Machine Learning
Data & Vector DBs
MLOps & Platform
Cloud & Infra
Credentials & Training
Formal certifications and training that underpin the practical experience.
TOGAF 10
CertifiedFoundation + Practitioner
Neo4j Certified
CertifiedLLMs · Knowledge Graphs · Data Science
DP-100 Training
CompletedAzure Data Scientist Associate Course
Data Science Fundamentals
CertifiedLloyd's & University of Southampton
What I Write About
This site is where I think through AI architecture problems in public. No hype, no buzzwords—just practical thinking about building AI systems that work in the real world.
Let's Connect
Whether you want to discuss AI architecture, explore collaboration opportunities, or just say hello—I'd love to hear from you.