Overstone Spotlight
Chris Mann: Head of Data
Could you tell us about your professional journey and what drew you to the field of data at Overstone? How has your experience shaped the way you approach data and analytics today?
I started my career in a rigorous, safety-critical environment as a Nuclear Engineer for Rolls-Royce, where I managed multimillion-pound submarine maintenance projects. That experience drilled into me the absolute necessity of strict engineering standards and robust stakeholder management.
I then transitioned into enterprise consultancy at Deloitte as a Lead Data Scientist, where I designed and delivered large-scale machine learning platforms, including a trade promotion optimisation system on Microsoft Azure.
What drew me to Overstone was the unique opportunity to combine my 'heavy engineering' background with the agility of a start-up. Today, my approach to data is entirely shaped by this blend: I focus on enforcing production-grade Machine Learning Operations (MLOps) and reliability, while maintaining the speed and innovation necessary to scale a WealthTech business.
Overstone combines human expertise with machine learning. From your perspective, what are the biggest benefits and challenges of this human-in-the-loop approach, and how does it enhance the reliability of your predictions?
At Overstone, we rely heavily on our art team to review data structuring, insights, and valuations, which is crucial for building trust in art as a true financial asset.
The biggest benefit of this human-in-the-loop approach is the assurance of quality in a regulated, yet highly subjective, market. Art market data is noisy - the value of Picasso's works vary by six orders of magnitude, for example - so having high quality, correct data is vital to minimise noise.
That said, the primary challenge is the sheer time burden this work places on our art specialists. My goal is to scale that human expertise by automating the heavy lifting. This is a continuously improving process and one that Large Language Models (LLMs) and Vision-Language Models (VLMs) increasingly play a part in.
Generative AI and large language models (llms) are increasingly shaping the art and data landscape. How do you see LLMs being integrated into Overstone’s workflow, and what potential do you think they hold for transforming data analysis in the art market?
We have been keeping an eye on deep learning and GenAI for years and have already integrated GenAI into many aspects of our workflow, from using Cursor (an agentic software development program) to performing entity resolution or data structuring.
The ability of LLMs to reason about the meaning of text already enables things like researching and analysing historical and upcoming exhibitions at a scale that was previously impracticable.
Increasingly, the visual capabilities of VLMs is of particular interest - it is no longer inconceivable that AI could perform visual appraisals of artworks in the not-too-distant future, for example.
Like many areas of our lives, I expect GenAI to transform the art market. I think it is only a matter of time before an AI acts on behalf of a collector to buy a work at auction, especially with OpenClaw unleashed online already. Failing that, perhaps an AI will use Rent-a-Human AI to bid in a live auction!
...and finally, if any artwork could hang on your wall, which one would you choose and why?
After all this talk of GenAI, I thought it would be entertaining to ask an LLM to answer this question.
After telling it Chien Chung Wei is my favourite artist and a bit about myself, here's what it said: “I would choose a piece by Chien Chung Wei. Given my academic background in physics, I am naturally drawn to his structural mastery of light, shadow, and architectural composition. His work is incredibly precise and analytical in its foundation, yet it flows beautifully and effortlessly on the canvas." Clearly the LLM hallucination problem has not been solved yet.