Henry speaking at an event

Hi I'm Henry

I turn messy information into reliable decisions.

I’m a builder and researcher working across AI, data, and markets. I combine research discipline, market experience, and long-term systems thinking to build dependable systems for high-stakes problems.

What I build

Data foundations

I turn messy, real-world data into dependable foundations for machine-learning and analytical systems.

Agentic systems

I turn reasoning-heavy workflows into reliable layers of agents, reusable skills, and deterministic automation.

Decision tools

I combine broad information, market data, and structured reasoning into decisions I can test and act on.

Selected projects

RSS3 blockchain data infra

At RSS3, I led blockchain data-platform and infrastructure work that cut cloud costs by 40%, moved ETL and indexing workloads from TypeScript to Go, and preserved portability across cloud providers.

40% lower cloud cost

Five-to-three cloud unitsFive equal cloud units become three, making the 40 percent reduction countable.

RAMPVis public-health research

During my PhD, I used D3.js extensively for data-visualization research.

10+ publications · £1M+ grants · UK Global Talent

RAMPVis research trajectoryA compact parallel-coordinate ribbon summarizes publications, grants, and recognition.

RSSHub open-source network

I’m a core contributor to RSSHub, an open-source RSS network with more than 40,000 GitHub stars and thousands of global instances.

40,000+ GitHub stars

RSSHub chord diagramSix community segments exchange routes through an interconnected open-source network.

How I think

Reason from first principles

I define the real constraint, test assumptions, and choose the simplest system that can carry the work.

First-principles reductionSuccessive assumption layers are stripped away to expose one irreducible core.

Extract signal from data

I gather evidence broadly, test it carefully, and narrow the data toward insights that change decisions.

Evidence filteringBroad evidence enters a testing filter, irrelevant points are rejected, and one usable signal exits.

Build for compounding

I turn repeated work into knowledge, automation, and feedback loops that become more useful over time.

Compounding feedback loopRepeated feedback loops return more capability to the system, ending in progressively larger gains.

How I learn

Nonlinear education milestonesA three-step path moves from skipping a bachelor’s degree to an MSc with Distinction and a PhD defended during COVID-19.Skippedbachelor’sMSc ComputerScience DistinctionPhD defendedduring COVID-19

My route into computing was nonlinear: I skipped the bachelor’s, completed an MSc in Computer Science with Distinction, and defended my PhD during COVID-19.

That same self-directed instinct shapes how I keep learning: I choose my feeds, self-host FreshRSS, contribute to RSSHub, and maintain an Obsidian knowledge base and information pipeline.

Now

Markets have been my laboratory since I started trading at 18. Today, research habits, data foundations, and automation converge in my AI system for reliable financial decisions.

Research habitsData foundationsAutomation systems
Convergence braidThree capabilities converge into reliable decisions.ReliabledecisionsConvergence braidThree capabilities converge into reliable decisions.ReliabledecisionsConvergence braidThree capabilities converge into reliable decisions.Reliabledecisions

Beyond the systems

Carrot the cat

Carrot the cat reminds me that his attention is all I need.

If you’re building AI systems that turn complex data into consequential decisions, especially in markets or data infrastructure, I’d like to compare notes.

Start a conversation (opens in a new tab)