Optiver

It is day two at @europython, and one of the questions we get most at the booth is whether we use AI. We do. AI has made writing code much faster, but as code becomes cheaper to produce, building codebases that are easy to review, maintain and evolve becomes even more important. Later today at 16:00, Software Engineer Dan Jones explores that in his talk, AI Slop to AI Gold: Building Codebases That Last. Until then, find us on the floor or learn more about tech at Optiver: optiver.com/what-we-do/tec…

Explaining Markets is live. Build an AI system for outcomes that are still unknown. Connect your agent, receive live earnings events, and track its performance on the leaderboard throughout Q3. Sign up by August 9: bit.ly/4526QMg

What does it take to move #AI from experimentation to enterprise scale? In our latest technology blog, Optiver engineers share perspectives alongside @databricks on the building blocks of successful enterprise AI. The discussion looks beyond foundation models to the broader ecosystem required for AI to succeed – including high-quality data, robust infrastructure, governance and platforms that can evolve as technology advances. Read the full blog: optiver.com/insights/techn…

Now live from 'Deep Learning for Trading', our exclusive @icmlconf event. Curious about how we apply ML in trading? Explore: optiver.com/what-we-do/res…

Why do investors use systematic internalisers? As policymakers consider reforms to Europe's equity markets, it's an important question. Understanding why investors choose different execution channels is crucial to designing effective regulation. SIs don't exist because investors want to avoid lit markets. They exist because they can deliver better outcomes for specific orders at a price that's better than the quoted spread while providing certainty of execution. These benefits are particularly valuable for institutional investors managing larger orders. We believe the recent ESMA call for evidence on European equity market structure overlooked two key measures of price improvement offered by SIs: size improvement and midpoint liquidity. Here, we show simplified examples of how these work in practice. As policymakers consider whether European equity market structure is fit for purpose, they should strive to preserve the competition that benefits investors while ensuring Europe's lit markets remain strong, transparent and resilient. The goal should be a market structure that delivers the best outcomes for investors, rather than one that favours one trading model over another. Read more here: optiver.com/insights/marke…

We're live for Day 2 at ICML! Visit our stand to meet Optiver researchers and engineers, and learn how we apply ML, quantitative modelling, and large-scale computing to solve complex problems across 100+ exchanges. [ICML] Int'l Conference on Machine Learning | #machinelearning #research #STEMcareers

As Europe considers reforming its equity market structure, one thing is clear: policy should be driven by evidence. A common narrative is that declining lit market volumes automatically mean weaker lit markets, but the data tell a more nuanced story. It's true that the share of trading executed on lit venues in Europe has declined, as shown in chart 1. But key measures of market quality – bid-ask spreads and top-of-book depth – have remained steady, as shown in charts 2 and 3. The question for policymakers shouldn't simply be where trading happens, but whether investors continue to receive competitive prices, deep liquidity and efficient execution on lit markets. And the evidence shows that this is happening. Read more here: optiver.com/insights/marke…

Gensyn at @ICML2026 Seoul July 6-11 Jul 8-9 connect x collaborate x cowork by Gensyn and @OptiverGlobal Jul 8 Pitch your Poster by Gensyn Jul 9 Research to Industry: Fireside Chat & Happy Hour by Gensyn and Optiver Jul 10-11 Four Gensyn paper presentations Thread below👇

Three days into ACCU @cpponsea and it's been great connecting with the C++ community. Our team has spent the week joining talks and connecting with engineers about performance, tooling, debugging, and the challenges of building complex C++ systems. If you're still here and we haven't crossed paths yet, stop by the Optiver booth and say hello.

Analytics agents promise "ask anything, get answers in seconds" and then deliver answers that are confidently wrong about half the time. The problem isn't the model. It's that nobody told the agent what the data means. On July 9, Optiver and @DataHubCloud are hosting an event in Chicago on what it takes to build a context layer your agents can actually trust. Expect a practical look at: • Why fragmented context breaks even the best models • How years of query history can become an immediately useful semantic layer • How Optiver uses DataHub and context intelligence to improve agent accuracy for trading analytics Learn more and RSVP: bit.ly/4xxdSFQ

Amsterdam’s capital markets are one of Europe’s great success stories – yet they rarely receive the attention they deserve. Our latest Insight examines how the Netherlands built one of the continent’s most competitive and liquid financial ecosystems: through deliberate infrastructure choices, a remarkably broad investor base and a dense cluster of liquidity providers committed to quoting prices in all market conditions. The Dutch story is also a timely one. As European policymakers work to build a deeper Savings and Investments Union, the conditions that made Amsterdam thrive offer a practical blueprint for scaling EU capital markets more broadly. optiver.com/insights/marke…

Most teams think of an agent as a standalone app, and think building many agents means building many standalone apps. On July 1, Matt Nassr, Head of Global Data and AI Transformation, is talking about how to build a shared substrate so every agentic workflow compounds on the last one. If you are attending AI Engineer World's Fair or based in San Francisco, seats are limited. Apply for a spot: gem.com/form?formID=a6…

As models become more capable, the focus is increasingly shifting from what they can do to how they perform and adapt beyond controlled environments. Excited to be back at #ICML2026, connecting with the global research community and exploring how advances in ML translate from research into real-world systems. Learn more: optiver.com/what-we-do/res… #MachineLearning #AI #Research | @icmlconf

When your platform’s users stop being people and start being agents, the job of a platform team changes. You’re no longer building for someone clicking through a UI. You’re building for an agent that needs context, guardrails, and the ability to act on its own. At @databricks Data + AI Summit, Calvin Yiu will share what Optiver’s Lakehouse team built to make that shift work: the infrastructure that sits between the data platform and the agents using it, which parts of Databricks they chose to build on and, just as honestly, where it still falls short. Learn more: bit.ly/4egl0xJ #DataAISummit

For years, policymakers have looked for ways to push trading activity back to lit markets. But measures like caps have been tried, and they didn't work. Instead of returning to lit markets, trading simply migrated elsewhere. Our new paper argues that Europe should stop trying to force the flow and start fixing the structural issues that put lit markets at a disadvantage. If Europe wants stronger equity markets, the answer is better market design. Here are 5 way to make that happen. optiver.com/insights/marke…

We're live for Day 2 at AI Engineer Melbourne. We run production trading systems across 100+ exchanges. Stop by to talk agentic AI, production systems and engineering at scale. @webdirections #aiengineer #aimelbourne #AIE #AI #machinelearning

When something feels slow, you’re probably looking in the wrong place. UI issues are rarely local. They’re usually caused by how data is fetched, recomputed, and duplicated across the system. Read more 👇 bit.ly/42aDEBg?utm_ca…

AI coding tools are moving quickly. The conversation is shifting from what's possible to what actually works in practice. As part of #NYTechWeek, Matt Nassr, Head of Global Data Engineering and AI Transformation at Optiver, will join leaders from @driver_ai_, @ShipBob and @LucaNet to discuss the workflow changes, trade-offs and organizational challenges that come with adopting AI coding agents. If you're interested in how engineering organizations are approaching these changes beyond demos and experimentation, it should be a worthwhile conversation. If you're in New York for NY Tech Week, registration is open while spots remain: bit.ly/4fgELY6

@Anthropic has released Claude Opus 4.8, and our headline finding is clear: performance continues to improve, especially at lower reasoning-effort settings. On Optiver's trading intern exam, Opus 4.8 delivers its strongest gains where less reasoning compute is required. That's an encouraging sign that frontier models are becoming both more capable and more efficient. The charts below show two views of that progress. The first tracks performance across twelve Claude model releases, from Claude 3 Haiku to Opus 4.8. The second compares Claude Opus 4.6, 4.7 and 4.8 across different reasoning-effort tiers. Read more about our research into #AI and #trading: optiver.com/insights/techn…

Teams building and scaling better agentic outcomes are building systems designed for agents to check their own work, and fix it, instead of relying on humans to detail corrections. At @aiDotEngineer Melbourne next week, Moss Ebeling will speak about what closed-loop design looks like and how you can use it to build real leverage via automation in your workflows. If you’re attending, check out this AI Engineering track event on Wednesday at 1:30pm and stop by the Optiver booth to talk with the team.