// HEAD OF AI
Jonathan Castro
AI Systems Architect
Designing and building production-grade AI systems and products. From agent architectures to RAG / Context pipelines—focused on what works at scale.
View on LinkedIn
View on GitHub
Head of AI at TopCode
(KSTF Group)
Former Co-Founder at The Cliff
(acquired)
University Professor
M.Sc. in AI
16+ years
building Tech systems and products
01. About
Core Philosophy
AI is not just about models—it’s about systems. The real challenge is building infrastructure that scales, orchestrates complexity, and delivers value in production.
Origins
My first steps in AI were 16 years ago, building game AI and real-time systems—simulation, decision trees, emergent behavior. That foundation in structured decision-making and autonomous agents shaped how I approach agent architectures today.
Evolution
16 years connected to AI across every era of the industry. Computer vision at Dive as Tech Lead. ML infrastructure at Hocelot as Tech PO. Technical product leadership at Eccocar. Then I co-founded The Cliff, an AI venture studio that I sold in 2025. Each stage built on systems thinking.
Present
Head of AI at TopCode (KSTF Group), the company that acquired my venture studio’s assets. Building production-grade AI systems for Tier 1 global banks—trade finance, customer support, and enterprise infrastructure at scale.
Academic
University professor teaching official M.Sc. in AI. PhD in progress. Bridging research and production—what works in theory vs. what works at scale. Developing open-source AI solutions and actively contributing to the community.
02. Current Focus
Building production AI systems that solve real problems at scale
Agent Architectures
Multi-agent orchestration with intelligent routing, tool use, and decision frameworks
SDD Frameworks
Structured development frameworks for building, testing, and deploying AI-powered applications at scale
AI Infrastructure
Observability, cost optimization, and production-grade deployments
Context and RAG Systems
Retrieval pipelines, hybrid search, and semantic chunking at scale
AI Product Systems
From prototype to production—building AI that ships and scales in real-world environments
Real-time & Voice AI
Low-latency systems for conversational interfaces
CUA: Computer Use Agents
Autonomous agents that interact with desktop environments, web browsers, and software interfaces to execute complex tasks
03. Selected Experience
Leading AI strategy, systems design and product development across multiple high-impact initiatives within the KSTF ecosystem, following the transition from my previous venture, The Cliff, and the integration of its core team.
Key Achievements
- ▸
AI team and technical direction: defining architecture, tooling and development practices for AI systems in production environments
- ▸
AI-native systems: designing and delivering agent-based architectures, RAG pipelines and LLM-driven workflows
- ▸
Trade Finance AI: driving AI initiatives across confidential projects for Tier 1 global banks
- ▸
AI-powered customer support: leading the development of systems for large-scale enterprise environments
- ▸
Systems × product × infrastructure: operating at the intersection, ensuring scalability, reliability and real-world impact
Technologies
AI venture studio focused on designing and building production-grade AI systems and products. From day one, we operated at the intersection of engineering, product and research, delivering high-impact solutions for companies across Europe and the Middle East.
Key Achievements
- ▸
Agentic systems and RAG architectures: specialised in AI-native software, synthetic data platforms and experimentation environments
- ▸
Digital twins and simulation: built solutions using Unreal Engine and Omniverse, including physical AI and real-world system integration
- ▸
Proprietary products: developed maind.pro, exploring new paradigms in AI-driven applications
- ▸
International delivery: collaborated with NVIDIA, Broadcom, iStyle, Nakheel, Mango, LaLiga and KSTF across Europe and the Middle East
- ▸
Strategic acquisition: in 2025, a key client acquired the core assets — including team and AI IP — marking the transition of the company
Technologies
Teach in an official graduate programme in Artificial Intelligence, focusing on bridging theory and real-world AI systems.
Key Achievements
- ▸
NLP and AI product innovation: deliver courses covering modern paradigms such as LLMs, RAG and agent-based architectures
- ▸
Theory to practice: translate complex concepts into practical system design frameworks, connecting academic foundations with production-grade AI applications
- ▸
Student mentorship: guide students in the development of AI systems and products, combining technical implementation with product thinking
- ▸
Industry experience: bring hands-on experience into the classroom, with a focus on moving from models to real-world AI systems
Technologies
Led product development for a mobility platform, focusing on scalable, API-driven systems and aligning product strategy with engineering execution.
Key Achievements
- ▸
Product vision and roadmap: translating business needs into technical requirements and system-level decisions
- ▸
System design: worked closely with the CTO and engineering teams on APIs and integration patterns (API-first)
- ▸
Cross-functional delivery: led collaboration across engineering, QA and DevOps, improving delivery pipelines, reliability and product quality
- ▸
Architecture and data flow: contributed to scalability and efficient system interactions
- ▸
Product discovery: conducted market analysis, informing both product and technical direction
Technologies
Led the design and delivery of data-driven products and AI systems in the insurance and financial sectors, working at the intersection of product, data science and engineering.
Key Achievements
- ▸
Product vision: translated business requirements into scalable system architectures, enabling data and ML-driven capabilities
- ▸
End-to-end ML solutions: worked closely with Data Science, Engineering and DevOps teams to design and deliver ML-based solutions
- ▸
API-first approach: defined APIs and system interfaces ensuring integration across distributed systems
- ▸
ATDD practices: established structured workflows, improving delivery quality and alignment between product and engineering
Technologies
First employee and key contributor in the development and scaling of an AI-driven platform leveraging computer vision to extract and enrich real-time content from films and TV. Started as the first technical hire and evolved into a technical product leadership role as the company scaled.
Key Achievements
- ▸
Prototype to product: led the transition, defining architecture, tech stack and workflows across mobile and Smart TV platforms
- ▸
Engineering organisation: built and scaled from an early-stage setup to ~75 people
- ▸
Technical and product decisions: drove decisions aligning system design with business goals and product strategy
- ▸
Cross-functional management: managed teams across mobile and interactive platforms, delivering high-impact features in fast iteration cycles
- ▸
Hands-on prototyping: developed prototypes and key components across Unity, HTML5 and mobile technologies
Technologies
Founded BigKahuna Games while at university and worked at Bitoon Games, building videogames and real-time interactive systems from scratch. Early experience building real-time decision and simulation systems that closely resemble the foundations of modern agent architectures.
Key Achievements
- ▸
Game AI systems: designed and implemented FSMs, behaviour trees and fuzzy logic, modelling structured decision-making for NPCs
- ▸
Pathfinding and navigation: developed A*, graph-based algorithms and nav meshes, working with spatial reasoning and dynamic environments
- ▸
Real-time physics systems: built systems combining physics, ray tracing and advanced vector/matrix operations under strict performance constraints
- ▸
Early VR exploration: explored Oculus Rift environments at Bitoon, working with immersive interaction and real-time simulation
- ▸
Internal tools: built tools and pipelines improving content production and development efficiency
Technologies
Liminal - Last Posts
View all posts
Exploring AI systems, architectures and ideas at the edge of change. Where what works today is already becoming obsolete.
From the Latin līmen, meaning threshold — the point between two states. In technology, we are always in that moment.
2026-04-20•10 min readComing Soon: Deep Dive into Production AI
An in-depth exploration of building and scaling AI systems in real-world production environments.
2026-04-18•12 min readComing Soon: Advanced Agent Patterns
Exploring cutting-edge patterns and architectures for building sophisticated AI agent systems.
2026-04-16•15 min readComing Soon: ML Infrastructure Best Practices
Lessons learned from building scalable machine learning infrastructure at enterprise scale.
05. Get in Touch
Open to AI leadership roles, advisory positions, and speaking opportunities. Let’s talk about building AI systems that actually work in production.
jonathan.castro.miguel@gmail.com
github.com/JonathanCastroMiguel
View GitHub Profile