Dorunexaari
Explainable regime insight
How we think about AI and volatility regimes
The usual story goes like this: someone drops a complex model on your desk,
wraps it in a sleek interface and calls it insight. When the next volatility
shock arrives, the same model becomes a mystery that no one wants to own. We
built Dorunexaari to avoid that pattern. Our background sits at the intersection of
quantitative research, risk oversight and data engineering. Over time we kept
seeing the same gap. Teams could describe today’s volatility term structure in
detail, but struggled to track when regimes were quietly shifting across
maturities and asset classes. By the time the change was obvious, the room for
careful adjustment had narrowed. Our response was a simple internal method we
call Regime Lens. It is not a magic formula. It is a repeatable way to structure
data, run AI models and frame outputs so that a risk committee can ask hard
questions and get clear answers. Each step is documented, from data filters to
model choices and validation checks. We focus on Canadian market participants
and global teams who must respect Canadian rules. That means we think about
governance first. We design workflows that make it easier to show how a signal
was produced, who reviewed it and where its limits sit. We would rather surface
model uncertainty than pretend it does not exist. Along the way we have learned
that small practical details matter more than grand promises. A well timed alert
about a potential regime change in the volatility curve, paired with a short
plain language note, can support better internal conversations than a thick
technical report that arrives too late. Our aim is to make those moments of
clarity more frequent without overselling what AI can do. We keep our research
agenda active, test methods on varied market conditions and refine the system as
new data arrives. Past performance does not guarantee future results, and we
design every feature with that fact firmly in mind. Results may vary, and
thoughtful human review always stays in the loop.
How our approach supports your governance responsibilities
Our role is to turn complex AI driven regime detection into plain, defensible inputs for your financial market research and risk governance, without pretending that models can see every twist in future volatility.
The easy story says that once you plug AI into your data, the hard parts of volatility analysis will take care of themselves. Our story is different. We stay close to the difficult bits, where regimes change slowly, explanations matter and decisions are shared across teams.
We built our practice around real conversations with risk officers, researchers and data teams. Many described the same pattern. During calm periods, detailed volatility reports felt like more noise than help. During stress, those same reports felt too slow or too opaque. That tension guided our design choices. We asked what kind of AI support would be welcome in both states. The answer was not more charts. It was a small, dependable layer of regime aware insight that could sit behind existing processes and documents.
Our internal methods borrow from machine learning, time series analysis and practical risk management experience. We treat models as draft storytellers. They propose where regimes might start and end across the volatility curve. We then refine those stories through validation, stress checks and clear documentation. The end result is not a single forecast, but a structured view of how current conditions compare with past clusters and where uncertainty remains high.
Because we work primarily with enterprise and institutional teams, we design for
committees, not individuals. That means clear audit trails, language that travels well
between technical and non technical readers and an honest account of limitations. We do
not frame our tools as a replacement for human judgment. Instead we aim to give people
better raw material for their own debates about risk, allocation and scenario planning.
Past performance does not guarantee future results, and our work is built around that
simple reminder.
We care less about impressive model names and more about how AI regime signals behave in real risk conversations, under real time pressure, with real oversight questions on the table.
The people and practice behind our volatility regime work
Behind Dorunexaari is a small team that has spent many hours in rooms where volatility, models and governance collide. We bring that experience into how we design, test and explain every part of our AI volatility regime analysis workflow.
Our team blends quantitative research, risk oversight and data engineering backgrounds. Some of us have built volatility surfaces and stress tests; others have sat in review meetings asking difficult questions about them. That mix keeps our work grounded. We know that a neat chart is not enough. People need to understand where the signal came from, how it behaves under different conditions and what it does not cover.
We follow a simple method we call the Three Window Review. First we look at how a
proposed regime signal behaves across different historical periods. Then we test how
it holds up when data quality shifts or gaps appear. Finally we review how easily a
non specialist could read and discuss the output in a governance setting. A signal
that fails any of these windows does not move forward, no matter how elegant the
underlying model may be.
We also keep a regular research cycle focused on new approaches to regime detection,
explainability and uncertainty measurement. When we add or adjust a method, we
document why, how it changes behaviour and where it should be used with extra care.
Results may vary, and we encourage teams to treat every output as one input among
several, not a single source of truth. That attitude, more than any single
technique, defines who we are.
Why we built Dorunexaari for risk teams
From opaque models to explainable volatility regime insights
There is a popular myth that more data and more complex models automatically lead to
better risk awareness. In our experience, they often just lead to longer meetings.
We started Dorunexaari after seeing careful teams spend hours debating outputs they did
not fully trust, even though the charts looked impressive. Our approach stays narrow
on purpose. We focus on AI support for volatility regime analysis, especially around
term structures, where shifts can creep in quietly. Instead of chasing every
possible signal, we care about a smaller set of views that play well with existing
market research, scenario work and governance processes. We use a framework we call
Transparent Regime Stack. First we stabilise the data, so inputs reflect what teams
already recognise from their own sources. Then we run a combination of statistical
and machine learning methods that are chosen for clarity as much as performance.
Finally we package outputs as structured notes, not just charts, so people can read,
question and annotate them. We work with risk and research leads who answer to
boards, regulators and internal audit. That shapes our culture. We write in plain
English, document trade offs and highlight where the model is less confident. We
prefer an honest, slightly conservative signal over a bold claim that might not hold
when conditions change. Our team includes people who have sat on both sides of the
table: building models and challenging them. That mix keeps our work grounded in day
to day practice rather than abstract theory. We keep asking one question of every
feature we design: will this make the next difficult risk conversation a little
clearer for the people in the room.
Our guiding principles for explainable AI volatility regime analysis
Radical transparency in data and modelling choices
We design our AI volatility regime analysis so that a careful reader can follow the trail from raw data to final note. That means clear descriptions of data sources, filters, model choices and validation checks. When a potential regime shift appears, we want risk and research teams to see not just the headline, but the reasoning that sits behind it. This level of transparency takes more effort, yet it is the only way we have found to support real accountability in financial market research.
Human judgment remains at the centre of decisions
We treat AI as a careful assistant to human judgment, not a substitute for it. Our methods are built to suggest where volatility regimes may be changing, then step back and let people decide what that means for their own context. We avoid language that overstates certainty and we highlight where models are less confident. This keeps responsibility where it belongs, with the teams who understand their mandates, constraints and risk appetite best.
Built to complement existing risk workflows
We believe that good tools should fit into the rhythms teams already use to manage risk. That is why we focus on regime aware views that plug into existing reports, committees and review cycles instead of asking people to rebuild their world around a new platform. We would rather deliver one extra layer of clarity on volatility term structures than attempt to reshape every part of a team’s process in one move.
Continuous review as a core design principle
Markets change, models age and what worked last year may not hold under new
conditions. We keep our methods under ongoing review, with an eye on both
performance and explainability. When we adjust an approach, we share what
changed and why, including any new limitations. Past performance does not
guarantee future results, and our philosophy is to treat that as a design
principle rather than a footnote.
Where Dorunexaari fits inside your risk and research ecosystem
We exist for teams who want AI support on volatility term structures without surrendering judgment, governance or the ability to explain their decisions to non technical colleagues and oversight bodies.
Many teams meet us after trying heavy platforms that promised simple answers to complex
volatility questions. We prefer modest claims, careful methods and tools that sit
quietly inside existing risk and research routines rather than trying to replace them
outright.
In our early work we saw how quickly models can drift away from the questions that
matter to decision makers. A term structure view might look smooth, yet hide important
changes in shorter maturities or specific market segments. We began to focus on regime
detection because it gives teams a language for these shifts that feels concrete.
Instead of arguing about whether conditions are normal or extreme, people can discuss
how current readings compare with past clusters and what that might mean for their
planning horizon.
We build our tools around a simple rhythm. Data is checked and shaped in ways that are transparent. Models run on top of that foundation with clear logs and validation notes. Outputs arrive as a small set of views and comments, which can feed into existing reports and committee packs. Nothing about this is flashy. The value comes from the way it helps teams move from vague concern about market stress toward a shared, documented view of potential volatility regimes.
Over time, clients have told us that the greatest benefit is not a specific chart, but
the way the work changes internal conversations. Analysts feel more comfortable raising
early signals because they can point to a structured process. Senior leaders gain a
clearer sense of how much confidence to place in each signal because uncertainty is
surfaced instead of smoothed away. This is the quiet space where AI can support
financial market research without overstepping its role.
Our philosophy on AI volatility regime analysis
A common belief says that any AI model that fits history tightly must be the right tool for future volatility regimes. We have watched that belief fail often. Our philosophy is to treat AI as a disciplined assistant, not an oracle, and to design every step so that risk and research teams stay firmly in charge of interpretation and action.
Clarity before complexity
We think AI should earn its place at the table by making complex volatility term structures easier to discuss, not harder. That means we avoid black box choices where small parameter tweaks change outputs in ways that even specialists find hard to track. Instead we use methods where we can explain, in plain language, why a regime boundary moved, which data points mattered most and how sensitive the result is to underlying assumptions.
About Dorunexaari and our work
Most teams start with shiny AI dashboards and hope the math quietly behaves in the
background. We went the other way. We started with the messy reality of regime
shifts in volatility term structures and then asked where machine learning can help
without hiding the risk. Our work sits between research, risk and data teams who
need signals they can explain to a committee, not just to a model card.
We design AI workflows that help market researchers and risk leaders see volatility
regimes earlier, compare scenarios faster and document decisions in a way that
stands up to internal review. The tools stay practical, the language stays plain and
every output can be traced back to its inputs and modelling choices.
