Oliver Smith is a London-based model validation specialist with a focus on machine
learning applications in volatility forecasting. We discussed the realities of model
auditing, the necessity of clear communication, and the evolving demands of
analytics teams. His experience highlights the importance of documentation and
adaptability in the face of ongoing change.
Walk us through a typical day as a model validation specialist in London.
Most mornings begin with a backlog of model performance reports and flagged
exceptions. My job is to audit every aspect of volatility forecasts powered by
machine learning—inputs, assumptions, and outcomes. I’m not just looking for
compliance, but for the story each result tells. The hardest part is keeping up
with constant model changes while ensuring everything is documented and
understood by more than just the developer.
What are the highlights and main difficulties in your current role?
The best part is uncovering the hidden assumptions that even experienced
developers sometimes miss. When you find a subtle issue before it impacts live
forecasts, it feels like you’ve actually done your job. The challenge is
translating technical findings into actionable recommendations for non-technical
audiences. It’s a constant balance between detail and clarity.
How has machine learning impacted your validation work compared to previous
approaches?
Machine learning has made the validation process more complex—models are less
transparent, and error sources can be harder to trace. We now require more
robust documentation and frequent cross-team reviews. Traditional models felt
predictable; these, not so much. It keeps things interesting, but there’s more
room for error.
What should someone new to validation and auditing keep in mind?
My advice: learn to communicate technical issues simply. It’s not enough to spot
a problem—you have to explain its impact. Don’t get attached to being right. Be
ready to adapt your process as both technology and regulations evolve.
How has your perspective on validation analytics changed over time?
Early in my career, I thought success meant finding fewer problems. Now I know
that real value comes from documenting everything—especially the issues. The
field is moving so quickly that continuous learning is the only way to stay
relevant.
We use cookies and similar technologies to enhance site functionality, analyze
usage, and remember your preferences. For details, review our Cookie Policy or
adjust your settings below.
Essential cookies
Essential cookies are required for secure site access and basic functions.
Analytics cookies
Analytics cookies help us monitor usage and improve services. Data collected is
anonymized.
Marketing cookies
Marketing cookies may personalize content or advertising through trusted
partners.
