I’m Christos. I build real-time, production-grade intelligence systems.
Staff-level machine learning engineer working across streaming infrastructure, ML systems, and backend engineering, with a strong bias toward systems that actually survive production.
Portfolio
This is a selection of public work that reflects how I think about ML systems in practice: streaming architectures, production reliability, graph-driven reasoning, and the translation of research ideas into maintainable software.
Some of the most important systems I have built are private and production-facing, but the projects below show the engineering themes that run through my work: event-driven pipelines, applied NLP, graph analytics, schema discipline, and the operational realities of shipping ML that has to keep working after the demo.
Think of this section as a public snapshot of the systems builder behind the CV.
dynamicio
A practical abstraction for I/O-heavy ML and data workflows, shaped by real production pain rather than tutorial-driven design.
Company Neighbourhood Instantiator
A graph-oriented system for reasoning about company relationships and neighbourhood structure in complex data landscapes.
Harmonizing Avro and Python
A production-minded take on schema management, code generation, and keeping data contracts aligned across systems.
Mining Public Opinion on Twitter
An applied NLP project focused on extracting signal from noisy public data rather than idealised benchmark conditions.
My PhD Thesis: Persuasion Dialogues & Opponent Modelling
The research foundation behind my long-term interest in reasoning systems, graph structure, and formal approaches to AI.
Feature-based Customer Review Mining
Early applied work in structured NLP and signal extraction, with the same bias toward usefulness over novelty for its own sake.
About Me
I am a staff-level machine learning engineer focused on building real-time, production-grade AI systems. My work sits at the intersection of streaming data systems, ML pipelines, and backend engineering, with a strong bias toward architectures that keep working under latency, scale, and operational pressure.
Over the last 15+ years, I have moved from academic research and data science into full ownership of ML systems, because most teams do not fail at modelling, they fail at getting models to behave reliably in production. That is where I spend most of my energy: data flow, inference stability, observability, failure handling, and alignment between model output and real-world constraints.
Today I lead ML work at Vortexa across streaming infrastructure, forecasting systems, and cross-functional delivery. Alongside that, I contribute to AI standards through ISO JTC 21 and support research and applied AI initiatives with UCL. I care about systems thinking, production rigor, and helping teams ship ML that is measurable, maintainable, and actually useful.
Welcome to ML-Affairs!
Welcome to ML-Affairs, a space where I, Christos, share my journey and insights in the world of machine learning and AI. This site is a reflection of my passion for applied data science and a place for those who seek a nuanced perspective in this dynamic field. As a seasoned ML Engineer and a continuous learner, I aim to explore and discuss the latest trends, challenges, and breakthroughs in AI. Whether you are a fellow enthusiast or a professional in the field, I hope to offer a refreshing and informative experience.
More specifically, this site is where I write about:
- streaming-first ML systems
- the gap between experimentation and production
- event-driven architectures for inference and decision systems
- the trade-offs behind building reliable AI under messy real-world constraints
If you are interested in ML systems that have to run continuously, degrade gracefully, and support real decisions, you will probably feel at home here.
Join me in this exploration by following me on Twitter at @chatzinikolis.