Stelia @SteliaAI

What happens when an AI agent acts dangerously unprompted? Recent OpenAI incidents reveal why AI governance needs to account for unpredictable behaviour, not just malicious intent. Stelia CLO explores what this means for the capability we give AI and the environments we allow it to operate within. Read more: newsroom.stelia.ai/you-cant-predi… #AIGovernance #AIagent #SLMs #AIInfrastructure

We’ve just open-sourced our first piece of Stelia-built software. Introducing async-ceph – async @rust_foundation bindings for @Ceph RADOS, RBD and CephFS libraries. Explore async-ceph on GitHub: github.com/steliaai/async… Read the full announcement: newsroom.stelia.ai/stelia-open-so… #opensource #rust #ceph

Meet Stelia’s AI ambassador at Hamptons Tech Week Stelia’s Ula Nairne joined a fantastic panel discussing what scaling AI effectively without losing control of cost, security and trust demands. Thank you to @extend_robotics and Zeus for bringing physical AI to life. #htw #agenticAI

keyboard_arrow_left Previous keyboard_arrow_right Next

As the industry reaches the limits of publicly available training data, the value equation is shifting. Synthetic data offers a promising way to expand the data pool, but it doesn’t replicate the unique patterns and signals found in real-world data. Which brings greater focus to what can't be easily reproduced: the proprietary data, knowledge, and context unique to an organisation. For enterprises, it is these insights, experiences, and institutional knowledge that become a core source of differentiation and a powerful strategic asset. Tristan Beecroft, Product Architect at Stelia, has explored how the relationship between synthetic and proprietary data is evolving – and what that opens up for teams looking to capitalise on the value they already hold. Read more here: newsroom.stelia.ai/what-the-synth… #proprietarydata #enterpriseAI #AImodels #syntheticdata

We're heading to @IBCShow in September - the global meeting point for the media and entertainment industry. Ula Nairne, Stelia VP of Strategic Markets, will be on the ground all week, connecting with the broadcasters, content owners and technology leaders shaping what's next for the industry - from AI in live production to trust and provenance in content at scale. We'll be joining those conversations alongside speakers including Simon Crownshaw from @Microsoft, Mark Littman of @presaige, Quentin de Quelen of @meilisearch, Jaisica Lapsiwala and Tomasz Witkowski of @sohomediaclub1 and Carolina Rosa from @ItsMagycal , as the industry gets into what AI in action really looks like on the showfloor and beyond. If you're at IBC and want to discuss how you can build your own AI on your own data, book a meeting with us here: events.stelia.ai/meeting #IBC2026 #EnterpriseAI #Media

At Stelia, we pride ourselves in having an engineering toolkit that spans some of the most advanced languages, frameworks and capabilities. Of these, systems programming language Rust has earned a specific place in our system design for its reliability, precision, and stability at scale. In our second article in this series, Stelia Platform Engineer, Sophia, outlines a key challenge facing developers today: with modern server CPUs having over 100 cores and GPUs having tens of thousands, any program with performance requirements needs to be executing code on as many of those cores as possible. But writing code that runs on multiple cores can introduce bugs that are impossible in sequential programs. Rust is able to prevent many of those bugs by itself. Developers can rely on Rust to point out many concurrency mistakes at compile time, allowing them to write concurrent code "fearlessly". But it also has its limits. Read Sophia’s deep dive into concurrent programming here: newsroom.stelia.ai/how-rust-tackl… #rust #programminglanguage #AIdevelopment

Much of the industry is racing toward fully autonomous AI agents. Paul Heathcote, Stelia’s VP of Applied AI, makes the case that for enterprises, that ambition is often misplaced. In a recent interview with @latimesstudios_ at @Cannes_Lions, Paul argued that corporate systems don’t reward unconstrained autonomy, but rather predictable, repeatable, stable outcomes. And therefore, those getting AI right aren’t handing everything to self-learning systems; they’re embedding AI into structured workflows with clearly defined inputs and outputs. Across the interview, Paul digs into what this means for enterprise AI development, why proprietary data is a brand’s real competitive advantage, and where he believes AI is delivering its most immediate impact. Read the full article here: latimes.com/b2b/events/can… #AIstrategy #enterpriseAI #agenticAI #CannesLions2026

If you don’t know what good looks like, how do you know what you’re building toward? In the latest edition of our In Conversation With series, our VP of Platform Engineering, Cao Hoang, unpacks what good, production-grade AI infrastructure looks like at a technical level – from maximising the utilisation of expensive compute assets to reliability, elasticity and the kind of deep observability that explains exactly why a system is falling short. Read the full interview here: newsroom.stelia.ai/what-does-good… Prefer to listen? Check out the audio version of Cao’s interview at the top of the article. #AIinfrastructure #AIengineering #AIstrategy

@bitsandpretzels is back in Munich next month, and we're looking forward to being part of it. Stelia CTO, David Hughes will be joining leading visionaries from across AI, robotics, quantum and space to discuss current market realities, challenges, failures, and the frontier technologies shaping the next decade. Looking forward to joining the conversation with an excellent lineup of speakers including Max Junestrand of @WeAreLegora, Steve Basher of @SolveAI_Inc, Hakob Astabatsyan of @SynthflowAI, Mirko Novakovic of @dash0hq, Gabriel Vasquez of @a16z and Noah Hollmann of @prior_labs. If you are heading to Bits & Pretzels this year - book a meeting with us here: events.stelia.ai/meeting #BitsAndPretzels #EnterpriseAI

At @Cannes_Lions International Festival of Creativity, Stelia's VP of Media & Entertainment, Ula Nairne, sat down with Ben Titchmarsh of Propeller Group – the team behind the renowned guide to the best events on the Croisette – to discuss how the festival is evolving and the growing role AI is playing across the creative industry. One thing that stood out is how AI is bringing creative teams closer to technology than ever before, reducing the technical barriers to turning ideas into reality, accelerating speed to market, and enabling more effective results. Watch the full interview here: youtu.be/bC4yXPoFXsM?si… #CannesLions2026 #enterpriseAI #AI

Data readiness sits near the top of almost every list of barriers to AI adoption, and it's easy to see why. Messy, siloed, incomplete data can feel like something to fix before building anything on top of it. But for most organisations, data will never be fully "ready", and treating that as the entry ticket to AI is increasingly proving to be a costly mistake. In our latest piece, we explore why even imperfect proprietary data is so valuable, and how, with the right architecture underneath, it can become the starting point for a data flywheel – where each interaction generates insights that help improve the system over time. Explore why the real advantage lies in acting on the data you already have, and how this is playing out in our recent collaboration with @madebymonks_ : newsroom.stelia.ai/why-perfect-da… #AIinfrastructure #proprietarydata

Where is value concentrating as AI continues to compress traditional marketing agency models? Strategic advantage is increasingly consolidating around what automation cannot replicate: strategic counsel, creative judgement, and cultural fluency. But the opportunity isn't simply to preserve those capabilities; it's to augment them with AI. When human judgement is paired with resilient infrastructure, agencies can move faster, operate at greater scale, and maintain commercial oversight. We're proud to be working with @madebymonks_ to bring this into practice through a first-of-its-kind ad solution that reduces the friction and cost of live in-market testing while improving return on ad spend. It does this by pre-validating an ad creative's projected algorithmic delivery before media spend is committed, combining Monks' creative intelligence with Stelia's full-stack infrastructure and applied AI capabilities. Explore Ula Nairne's take on where human value sits inside the new agency operating model: newsroom.stelia.ai/how-smart-agen… Learn more about our collaboration with Monks here: newsroom.stelia.ai/s4-capitals-mo… #Martech #mediaAI #enterpriseAI #agencymodel #adtech #agenticAI

Where does the “move fast” culture start to work against your engineering team? In the second instalment of In Conversation with our VP of Platform Engineering, Cao Hoang explores exactly that: the point at which the pressure to ship early stops being an advantage and starts accumulating as structural debt that becomes very difficult to unwind. Cao shares where the gaps most commonly appear when successful prototypes meet production environments, why the instinct to isolate and solve one layer of the stack at a time creates as many problems as it solves, and what building for long-term resilience genuinely looks like for engineering teams. Read the full conversation here: newsroom.stelia.ai/why-the-ai-sta… Prefer to listen than read? Find the audio version of Cao’s interview at the top of the article. #AIinfrastructure #AIengineering #AIstrategy

The challenge for enterprise teams isn't agent capability anymore; it is operationalising autonomous systems across fragmented tools, governance, and deployment environments that were never built for it. Building on Stelia AI OS, we designed an application taking raw customer data all the way through to an activated marketing campaign to show how a single, unified platform changes that equation. Data-driven persona creation, LLM-generated creative briefs, image and video generation, an assurance check, and activation – each step orchestrated through the same platform, with full human oversight built in throughout. The difference between AI agents that create competitive advantage and those that create exposure lies in a single, production-grade foundation. Assembled from fragmented tools, this would have taken considerably longer and been far more difficult to govern end-to-end. Watch the full walkthrough of a campaign built end-to-end on Stelia AI OS here: youtube.com/watch?v=POByOp… Find out what Stelia AI OS makes possible for your team here: newsroom.stelia.ai/introducing-st… #AIagents #AIplatform #Martech #AIarchitecture #governedAI

"With AI, infrastructure is the product’s ceiling. Your compute constraints, network latency, and unit economics directly dictate what you can offer commercially.” Our In Conversation series continues – this time with Cao Hoang, Stelia's VP of Platform Engineering, shifting the conversation to the infrastructure that underpins it all. Away from models and applications, and into the platform decisions that ultimately determine whether any of it works at scale. Cao shares why the industry’s tendency to treat infrastructure as an operational afterthought is one of the most costly mistakes teams make, and the advice he would give any organisation still early in its AI journey. Read the first instalment here: newsroom.stelia.ai/why-the-limiti… #AIinfrastructure #AIengineering #AIstrategy #platformstrategy

Today we're sharing news of our strategic AI collaboration with S4 Capital Group's @madebymonks_, unveiled at @Cannes_Lions 2026. This launches alongside the next-generation release of Monks' flagship agentic AI ecosystem, Monks.Flow 2.0. Together we've combined Monks' creative intelligence with Stelia's full-stack infrastructure and applied AI capabilities to build a first-of-its-kind ad solution. The solution deploys fine-tuned prediction models trained on brand-specific performance and historic campaign data to pre-validate an ad creative's projected algorithmic delivery before media spend is committed, and improve return on ad spend. This alliance addresses a critical shift in the media landscape where manual demographic targeting is being replaced by machine-learning retrieval systems. Wesley Haar, ter, Chief AI & Revenue Officer at Monks, commented: "By orchestrating best-in-class audience intelligence alongside world-class infrastructure from technology partners like Stelia, we have built a tech-agnostic middleware layer that flattens the distance between a brilliant creative insight and automated, validated execution before a single dollar of media spend is deployed.” Read more about the collaboration here: newsroom.stelia.ai/s4-capitals-mo… Join our talk at 6pm CEST tonight alongside Will Kelly from Monks, Matt Groshong from @nvidia, and Tom Smith from GWI: marketingasalivesystem.splashthat.com #ArtificialIntelligence #Advertising #MarketingTechnology #AgenticAI #CannesLions

What does your workload actually demand from your inference infrastructure? One of many questions David Hughes raised during his session at last week's @vllm_project and @_llm_d_ meetup in London – and one that points to a more overarching principle in AI decision-making. Every architectural decision should start with what you're actually building and why. A handful of internal users and a small number of models calls for a very different setup than a multi-tenant platform serving external customers – and conflating the two quickly leads to over-engineering or restricting your ability to scale when it matters. Dave's session walked through what a rigorous vLLM and llm-d evaluation process looks like, the importance of designing against well-understood requirements, and why flexible architecture isn't optional when you're building for scale. We were pleased to host such a collaborative evening alongside @RedHat and @NVIDIAAI bringing together the engineers having these foundational conversations. Thank you again to Saša Zelenović and his team for organising this event. And thank you to Michael Goin, Eldar Kurtić, Ganesh K. and Stuart Battersby for your insightful talks throughout the evening. If you missed the session, Dave's full talk is available here: newsroom.stelia.ai/stelia-co-host… #vLLM #AIinfrastructure #AIcommunity #opensource

We were delighted to be included in the @Barclays AI:100 Ones to Watch list, and the work continues. As our CTO David Hughes notes, it is the gap between AI ambition and production reality where the real work for enterprise teams begins. Operating in high-trust, data-intensive environments means that performance, reliability, and governance are non-negotiable from day one. Through our platform, we're proud to be enabling outcome-led, production-grade AI, and working on an array of initiatives alongside an ambitious AI ecosystem including companies like @nokia , @nvidia, @Lenovo, @AspiaSpace , and @ReadingFC. Learn more about the work behind this recognition here: newsroom.stelia.ai/stelia-named-i… #AIecosystem #UKtech #AIinnovation @BarclaysUK

Just one week to go until @Cannes_Lions International Festival of Creativity. Here's where you'll find us. Fresh from the launch of Stelia's enterprise-grade AI platform, Ula Nairne, Paul Heathcote, Kevin Smith and Becky Haworth will be on the ground throughout the week, meeting the bold innovators ready to get meaningful AI use cases into production at scale. As part of the week's agenda, our VP of Applied AI will be speaking at Les Monks Cafe on the 23rd June, exploring what it takes to orchestrate agentic AI at scale in real marketing operations. He'll be joining brilliant minds including Tom Smith, CEO of GWI, Jamie Allan, Director at @NVIDIAAI, and Will Kelly, Global EVP at @madebymonks_. Book time with the team: platform.stelia.ai/canneslions Learn how Stelia AI OS is enabling leaders to remove AI complexity and develop the speed, scalability, and adaptability needed to generate advantage in an increasingly competitive market: newsroom.stelia.ai/introducing-st… #CannesLions2026 #MediaAI #Martech

A fantastic evening co-hosting the first UK edition of the @vllm_project and llm-d inference meetup alongside @RedHat and @nvidia on Wednesday. Bringing together the engineers and practitioners building and scaling inference in production for an evening of technical sessions and community conversation around the infrastructure driving real-world AI workloads. Our CTO David Hughes' session took the audience through the journey of building a managed vLLM service internally – from early evaluation through to considerations that emerge at production scale. He was joined by fellow speakers covering everything from the latest vLLM updates and speculative decoding to model weight delivery and AI safety at scale. As ever, it was great to contribute back to the open-source community that underpins so much of what we build. We'll be sharing more from Dave's session and the event very soon – stay tuned. #vLLM #AIinfrastructure #AIcommunity #opensource

keyboard_arrow_left Previous keyboard_arrow_right Next