How we build AI cross venue liquidity mapping for

Inside our work

Each image reflects a concrete step in our AI workflow, from raw venue feeds to tested liquidity maps that support research driven trading design without turning your stack upside down.

research team analysing cross venue liquidity heatmaps on large shared screens in a modern office — Reviewing liquidity maps

We build AI systems that keep the structure of the market visible while giving researchers a sharper view of cross venue liquidity behaviour.

Why we focus on AI cross venue liquidity mapping

We believe AI for financial market research should expose how liquidity behaves across venues, not hide that behaviour behind opaque scores or unexplained dashboards.

We built Nerqualyxis around a simple observation. Research teams were spending significant time stitching together partial views of liquidity from multiple venues, often with tools that were never designed for cross venue analysis. At the same time, many AI offerings skipped over the hard microstructure work and jumped straight to opaque signals. We decided to focus on the gap between these two extremes. Our work sits at the layer where data from different venues, instruments, and time horizons becomes a coherent, testable map that researchers can use as a foundation for their own ideas.

Our internal methodology, the Liquidity Topology Cycle, structures how we apply AI. We first construct a venue and instrument graph that encodes how trading can actually route through the market. We then estimate a liquidity surface over that graph using both direct observations and model based inference. Finally, we test this surface against realised execution outcomes to see where the mapping holds and where it fails. This cycle repeats, which means our models are always treated as provisional and subject to revision as markets evolve.

For teams designing research driven trading approaches, the benefit is straightforward. Instead of treating liquidity as a static backdrop, you gain a dynamic, venue aware layer that shows where depth concentrates, where it fragments, and how that pattern shifts over time. This does not prescribe any particular trading style. It simply gives your researchers a more accurate starting point for their own design decisions, while keeping the cost and complexity of AI grounded in documented, repeatable processes. Past performance does not guarantee future results, and results may vary.

We focus on one Nerqualyxis, AI for cross venue liquidity mapping, and we build our process around research teams that design and test trading ideas on top of that map.

We start every engagement with what we call a Market Structure Baseline. We document how your current systems see venues, instruments, and order types, and we identify where liquidity appears fragmented or opaque. This is not a sales audit. It is a technical inventory that exposes mismatches between how your tools describe the market and how the market actually behaves. From there, we define which parts of the venue graph matter most for your research use cases, and we avoid modelling everything at once.

Once the baseline is clear, we move into the Liquidity Mapping Build. We assemble data from your existing feeds and archives, then we apply a mix of deterministic rules and AI methods to infer cross venue relationships. For example, we might align instruments that trade across multiple venues, estimate effective depth across that cluster, and detect when liquidity migrates from one venue to another. Throughout this phase we log every transformation so that your team can review, challenge, or replicate the work.

The final step is the Research Integration Loop. We deliver the liquidity maps, surfaces, and venue level indicators into your preferred research environment, then we watch how your analysts actually use them. We track which outputs feed into research driven trading design decisions and which remain unused. We then simplify, remove, or extend components based on real usage instead of abstract feature lists. This keeps the system lean and aligned with your budget and priorities.

Our approach to AI driven cross venue liquidity research

We keep our process narrow so that AI driven liquidity research stays explainable, budget aware, and aligned with how institutional teams actually work day to day.

We build AI systems for financial market research, with a narrow focus on cross venue liquidity mapping and forecasting. We care about one thing only, seeing where liquidity concentrates, fragments, and shifts across instruments and venues. We design tools for research driven trading design, not signal vending or automated execution promises.

We work with budget constraints as a design input, not an afterthought. We start with existing data flows, then decide where AI models add real signal instead of extra complexity. Our role is to translate fragmented venue data into stable research inputs your team can test, question, and adapt.

team reviewing cross venue liquidity maps on shared market research screens

How we think about AI for cross venue liquidity research

0 first

Market structure defines the modelling boundary

We start every project by mapping the market structure you already face, including venues, instruments, and routing constraints, before touching models. This keeps our AI aligned with how liquidity can actually form and move, and it prevents us from producing elegant but unusable abstractions that ignore real execution paths and venue rules.

0 second

Transparency is a non negotiable requirement

We design our cross venue liquidity mapping so that every transformation, assumption, and model decision can be explained to researchers, technology teams, and oversight functions. This reduces friction, simplifies sign off, and ensures that when the market changes, your team understands exactly which parts of the pipeline to revisit.

0 third

Cost discipline shapes every technical choice

We treat cost as a constraint that shapes our architecture, not a number to justify after the fact. We reuse existing data, tools, and infrastructure wherever practical, and we only introduce new components when they add clear value to your research driven trading design process and can be maintained without hidden overhead.

0 fourth

Continuous correction beats static certainty

We assume our understanding of liquidity will always be incomplete, so we build feedback loops that expose where our maps diverge from observed outcomes. This humility keeps us updating models, refining venue relationships, and removing complexity that does not improve your ability to reason about liquidity across instruments and venues.

We operate from Ireland and work with teams that need AI support for liquidity research, but still want full control over their own trading design decisions.

Our team brings together quantitative researchers, data engineers, and market structure specialists who have worked with institutional order flow, venue feeds, and archival datasets. We do not try to cover every area of financial research. Instead, we concentrate on cross venue liquidity mapping and forecasting, because that is where we saw the largest gap between available data and usable insight. This focus lets us refine our internal methods continuously rather than spreading attention across unrelated topics.

We maintain a simple collaboration model. We work directly with the people who own research questions, not only with procurement or technology intermediaries. We share working drafts of mappings, model diagnostics, and venue graphs early, so that your team can flag issues before they turn into technical debt. We also accept that some ideas will not survive contact with your real workflow. When that happens, we remove unused components instead of defending them.

We keep our infrastructure choices pragmatic. We reuse your existing logging, monitoring, and access control where possible, and we document what we add in plain language. This keeps operational risk visible and keeps the cost of running AI based liquidity mapping in line with its actual value to your research process. We aim for a setup that a new analyst can understand without needing to decode hidden pipelines.

Who we are and how we collaborate

We stay small, focused, and transparent so that research teams can see exactly how AI shapes their view of liquidity across venues and instruments.

Principles behind our liquidity mapping work

We organise our work around a few simple principles so that AI for cross venue liquidity mapping stays practical, auditable, and aligned with research driven trading design. We state each principle plainly, then we show how it changes day to day decisions about data, models, and cost control.

Microstructure first

We treat market microstructure as the starting constraint, not a detail to patch later. We map instruments, venues, and order types before we train any model, so the AI has a structure to respect. The result is a liquidity map that aligns with how orders actually route and fill, instead of a pattern pulled from unlabelled time series alone.

Transparent data work

We assume data is messy, incomplete, and expensive to clean. We prioritise transformations that can be described, repeated, and reversed. This keeps cross venue liquidity mapping explainable to risk, compliance, and technology teams, and it keeps maintenance effort predictable for your budget.

Models as hypotheses

We treat AI models as hypotheses about liquidity, not as facts. We expect them to drift, fail, and need revision. We track when and where their mapping disagrees with observed execution outcomes, then we adjust or replace them. This mindset avoids overconfidence and keeps researchers in control of conclusions.

Integrate, do not disrupt

We design our workflows to integrate with existing research stacks instead of replacing them. We expose liquidity surfaces, venue relationships, and anomaly markers in formats your analysts already use. This reduces switching costs and lets your team phase in AI driven mapping at its own pace.

quantitative team mapping liquidity research workflow on whiteboard

Focused, testable, venue aware AI

We started Nerqualyxis to solve a narrow problem we kept seeing in institutional research teams. Everyone had venue feeds, reference data, and archives, but almost nobody had a consistent view of where liquidity actually formed and moved across those venues. Each desk improvised its own process, usually with brittle spreadsheets and scripts that broke whenever market structure shifted. We decided to treat cross venue liquidity mapping as a core research product, not a side project. Our approach is simple. First, we define a stable representation of liquidity that can survive venue changes, symbol migrations, and new instrument listings. Then, we use AI methods to infer missing links, detect anomalies, and maintain that representation over time. The output is not a trading rule. It is a research layer your team can interrogate. We apply a method we call the Liquidity Topology Cycle. Map the venue graph, infer the liquidity surface, then pressure test it against real execution outcomes. Each cycle either confirms the current mapping or exposes where the model is wrong. This keeps the system honest and keeps your team in control of design decisions.