Interests
Not a pastime — an identity. Markets, venture ecosystems, organizations, language, sport, communities: every system I'm drawn to is noisy, adaptive, and full of people. The work, and the pleasure, are the same each time — separating what matters from what merely moves.
01 — Signal
Noise is cheap. Signal is earned — by theory, by method, and by sitting close enough to the system to hear it.
02 — Where I Look
The noise: millions of ticks, headlines, and half-beliefs repricing every second.
Markets are the densest signal-extraction problem there is. I build machine-learning strategies that listen for structure under the churn — and deploy them live, where the system grades your hypothesis without mercy.
Intraday ML strategy · AR 194% · Sharpe 4.32 · FedAnalysis: every Fed word, tone & gesture as data
The noise: thousands of startups all claiming to be the next wave.
Hype is a signal-jamming device. Using Anthropic's MCP release as a natural experiment across 4,532 AI startups, I test which "agent" ventures are a genuinely new paradigm — and which are the Gen-AI wave wearing a new name.
The noise: org charts, job titles, and what firms say they do.
Inside algorithmic-trading desks, the real system is invisible: who overrides the model, when, and why. Ethnography is signal-finding at human scale — sitting close enough to see where judgment hides in the loop.
Human–algorithm interaction for analytical creativity · PhD dissertation
The noise: everything everyone says; most of it strategic.
Text and speech leak intent. I study what persuades when the persuader is human versus AI, and what a central banker's phrasing reveals before the market has finished parsing the sentence.
The noise: form, luck, narrative, and a bouncing ball.
Football is a beautifully bounded complex system — adversarial, time-limited, endlessly surprising. BallSignal turns World Cup 2026 into a prediction problem: models against chaos, with the scoreline as referee.
The noise: feeds, channels, and infinite content competing for attention.
Growing LLM+ from five people to 7,000+ was an experiment in network signal: which formats, rituals and roles make a research community compound rather than churn. The answer is structure — fellowships, cafés, showcases.
03 — Off Duty
Even away from the screens, the habit persists. Reading literature aloud at Girton's Ridding Reading Prize — Asimov, Le Guin, Zephaniah — is signal-finding of another kind: interpretation, characterisation, diction, drawing the author's intent out of the text. And in a room of readers rather than models, it is a good reminder that the most interesting complex systems are still the human ones.