The Research and Applied AI Summit (RAAIS) is a community for entrepreneurs and researchers who accelerate the science and applications of AI technology. The 10th annual summit takes place on June 12th, 2026 in London. We are delighted to announce Nikolay Donets, Head of Machine Learning Engineering at Revolut, as a speaker.
At RAAIS we have a focus on translating cutting-edge technology and research into production-grade products for real-world problems.
The platform behind production AI at Revolut
Nikolay runs Machine Learning Engineering at Revolut, where his organisation builds the platform that supports every production AI system inside the company - from classical ML for fraud and personalisation, to time-series foundation models, to the voice agents now serving customer support. Revolut has crossed $1.3 trillion in transaction volumes and is the number one finance app in 19 countries; machine learning now sits in the path of millions of financial decisions a day.
The most concrete recent example of that platform in production is the rollout of voice agents across Revolut’s customer service operation, built with ElevenLabs. The system handles live calls in more than 30 languages, resolves tickets in under five minutes - roughly 8x faster than the previous escalation path - with a 99.7% call-handling success rate across more than four million customers in the UK and Europe.
One platform for builders, operators, researchers, and compliance
A central theme in Nikolay’s public work is that the hard problem in production AI is not building a model in isolation. It is building one platform that has to serve builders, operators, researchers, and compliance at the same time - and do so inside a regulated financial product. That framing is especially relevant now, because most organisations have already discovered that strong model performance does not by itself solve deployment. The harder challenge is the infrastructure around the model: evaluation, release discipline, governance, monitoring, and cost control, all without slowing iteration to a crawl.
For a technical audience, this is where a large share of the field’s practical difficulty now sits. As production AI moves into regulated settings - finance, healthcare, public services - the systems around the model have to satisfy operational and supervisory requirements as well as engineering ones. The platform is not separate from the model work. It is what determines whether model progress becomes durable capability inside a real organisation.
Governance as a velocity enabler, not a blocker
Nikolay has publicly outlined a framework for launching GenAI products in 90 days under regulatory constraints, built on three pillars: data lineage (treating compliance data as feature material rather than overhead), continuous delivery with multi-layered validation that goes beyond pass/fail tests, and compliance guardrails plus the documentation needed to defend them. The underlying claim is that governance, designed well, is a velocity enabler - moved into the development environment with clear tiers, predictable review cycles, and regulation treated as a technical requirement with a defined path to production.
As more companies try to support classical ML and generative AI side by side inside regulated products, this is becoming the central question in deployed AI. The bottleneck has shifted out of the model and into the systems that surround it.
Nikolay’s background
Nikolay holds a PhD in engineering from Siberian Transport University, where his thesis applied wavelet transform analysis to damage detection in beam superstructures from the response of traversing vehicles — structural health monitoring for bridges, an early grounding in reliability, monitoring, and operational discipline for critical infrastructure that carries through to his current work. His career has spanned Moscow, St Petersburg, Seoul, Stockholm, Toronto, and now London. He maintains active open-source projects and writes publicly on MLOps, AI governance, and risk in fintech at donets.org.
Short bio
Nikolay Donets is Head of Machine Learning Engineering at Revolut, where he leads the team that builds the AI platform behind the company’s production models - covering classical ML, time-series foundation models, and the voice agents now serving customers in 30+ languages. He has publicly outlined a 90-day framework for shipping GenAI products under regulatory constraints, built on data lineage, continuous delivery, and compliance guardrails. He holds a PhD in engineering, with earlier work in structural health monitoring and predictive maintenance for critical infrastructure.