For my assignment this week in my quantum course – as I cast around for a research topic for my second PhD – I was asked to write about where I hope quantum computing will go. My answer is pragmatic: commercial application will drive innovation, just as it always has. And when I think about innovation done well, I think immediately of Andrew Skirton, the best manager I ever had, who was the joint CEO of Barclays Global Investors before he retired (the Chief Investment Officer when I joined). The photo I printed here with his permission – taken last year at our annual Christmas gathering, decades after our team disbanded – is a reminder of how closely we as a team worked together.
Almost 30 years ago, long before machine learning became mainstream, I worked on portfolio optimisation in a quantitative investment fund. My role was to design algorithmic methods for constructing equity portfolios aligned with investor risk–return preferences. Under Andrew’s leadership, we were encouraged to think mathematically, structurally, and from first principles. Our models absorbed qualitative and semi‑structured inputs – analyst reports, macroeconomic commentary, news flow – and translated them into allocation decisions. It was, in retrospect, a primitive form of multimodal modelling.
Two constraints limited performance.
First, structured data was scarce. NLP barely existed. Sentiment analysis was embryonic. Extracting signals from unstructured text was computationally expensive and mathematically crude. Much of the information pipeline still depended on human interpretation — something Andrew insisted we treat as a strength, not a weakness.
Second, computational tractability was a major barrier. Real‑world portfolio optimisation is inherently high‑dimensional and combinatorial. Incorporating transaction costs, liquidity constraints, regulatory limits, dynamic correlations, and behavioural responses produces a non‑convex landscape that classical solvers struggle with. Even with heuristics, solving these problems in near‑real time was infeasible.
But Andrew cultivated a culture where constraints were not excuses. Our tagline — Engineering solutions for our clients — was not marketing; it was a mindset. He pushed us to innovate within the limits of classical computation, long before quantum computing was even a field.
And this is where the story loops back into quantum physics.
Quantum Computing as a Natural Extension of the Problems Andrew Taught Me to See
Quantum optimisation algorithms – QAOA, quantum annealing, variational quantum eigensolvers – are designed precisely for the kinds of combinatorial problems we wrestled with. They operate in a Hilbert space whose dimension grows exponentially with the number of qubits. A system of qubits spans a 2n-dimensional complex vector space, enabling the representation of entire allocation landscapes as amplitude distributions.
This is not just faster computation. It is a different mathematical substrate.
QAOA, for example, constructs a parameterised quantum state:
where encodes the cost function — the portfolio objective — and HM mixes the state space. The optimisation becomes the sculpting of interference patterns, not the traversal of a classical landscape. This is exactly the kind of structural thinking Andrew encouraged: understand the geometry of the problem, not just the mechanics.
Quantum annealing, similarly, uses adiabatic evolution:
keeping the system in the ground state as the Hamiltonian slowly transforms. It is mathematically elegant – and conceptually aligned with the adiabatic approximations I used in nuclear physics decades ago.
Andrew’s Influence, Seen Through a Quantum Lens
What I realise now is that Andrew Skirton modelled the very principles that quantum systems embody:
- Non‑linearity: He encouraged us to think beyond linear models and embrace complexity.
- Superposition: He taught us to hold multiple hypotheses simultaneously.
- Entanglement: He built teams whose thinking was correlated, aligned, and mutually reinforcing.
- Interference: He valued constructive challenge and the cancellation of bad ideas through debate.
In a sense, he ran the world’s first quantum‑inspired investment team — without any qubits.
Where This Leads Me Now
Quantum computing offers a pathway to the dream we had at BGI: real‑time, multi‑modal, multi‑factor optimisation that adapts continuously as information flows.
Quantum machine learning may finally allow us to integrate unstructured data — news, sentiment, macro signals — through entangled representations that capture correlations classical models cannot express.
And so, as I explore research topics for my second PhD, I keep returning to the same question:
How can quantum optimisation and quantum‑enhanced learning reshape real‑world decision systems — in finance, health, and beyond?
It feels like a full‑circle moment. The mathematics I learned in nuclear physics, the optimisation I practised under Andrew’s leadership, and the quantum tools emerging today are converging.
And perhaps that is the real lesson: Innovation is not a sudden leap. It is a long, coherent evolution — guided by the people who taught us how to think.
