It started with a Man United fan page, not a whitepaper.
I was the kid trading in-game items before I even understood what an economy was. Just obsessed with anything that let me transact, build a following, or run some kind of small operation of my own. By fourteen I was running a Manchester United fan page that grew to 30,000 followers, and because one obsession is never enough, a small e-commerce store on the side too.
At eighteen that turned into throwing nightlife events. Three of them cleared $30K+ in pure profit, which honestly isn't a bad early lesson in product-market fit, even when the product is a Friday night and the moat is knowing which fifty people to text first.
What actually pulled me toward crypto wasn't the charts. It was the same thing that pulled me toward all of it as a kid, a world with rules you can actually read, written in code instead of decided behind someone's desk. Web2 gave me the playground, I just kept wanting one with fewer landlords. That's a slightly dystopian way to see it, and I'll own that. Most interesting systems look a little dystopian right up until they don't.
Last year I did a summer school program at HEC Paris. I'm doing a distance learning degree with the University of London, so chances to actually go do something overseas are pretty rare, and I wasn't going to waste this one. I used most weekends there to go see things instead of just watching them on a screen, Parc des Princes on a whim, and Old Trafford, finally, after years of running that fan page from a bedroom on the other side of the world.
Side Quests
Grew a Manchester United fan page to 30,000 followers, and ran a small e-commerce store on the side.
$30K+ in profit across three self-run nightlife events.
Summer school at HEC Paris. Finally made it to Old Trafford in person.
Obsessed with systems where the rules are actually readable.
Three things building in web3 taught me the hard way.
Risk is a narrative problem before it's a math problem.
A composite score doesn't change anyone's mind. The story behind it does. At CryptaVision, the hardest part was never computing a risk number. It was explaining, in plain language, why a fund should care about a protocol dependency three hops away before it became tomorrow's headline. People act on stories they understand, not decimals they don't.
Primitives fail quietly, then all at once.
You can be recognized as a top builder in a category, like HAOO was in conviction markets, and still watch the category itself not show up on schedule. Markets don't reward being early. They reward being early to something that was always going to happen anyway. Telling the two apart in real time is the actual skill, and I didn't have it yet.
Trust doesn't scale by adding more dashboards.
Every risk or agent product I've touched starts the same way. Someone bolts a verification step onto the outside of a system that was never designed to be trusted from within. It works, until it doesn't. The interesting problem, the one I keep circling back to, is building trust in from the start, especially for agents that will soon be transacting with nobody watching the dashboard at all.
Where things stand today.
I studied finance, but these days almost all of my time goes into building. AI agent tooling, crypto infra, and whatever idea currently won't leave me alone. Right now that's an Associate seat at Protégé Ventures, scouting and backing student founders, plus building on Solana as part of Solana Fall School.
I don't have a fixed answer for where any of this leads, and I'd rather find out by shipping than by waiting around for clarity first. Outside of work I try to actually see the places I hear about, five countries in the past year and counting.
Currently
Solana Fall School, building on Solana and shipping in public.
The YC workshop at SQ Collective, and the Telegraph Protocol Hackathon, where I built vernier.
Occasional notes on crypto infra and risk, over on Medium.
Looking for the next thing worth building on-chain.
How the last few years actually went.
I scout and back student founders across the region, the earliest, rawest stage of the same builder path I'm on myself.
Trust infrastructure for autonomous agents in crypto markets. Still early, still mostly a thesis I'm testing by actually building it, more on that below.
I co-founded and bootstrapped a risk-intelligence engine for digital asset investors. We got real traction (45+ trading desks weighed in, including Pantera, Delphi, and Amber) before it just didn't work out. More on what we built below.
A few weeks with a small, fast-moving team. It's since gone on to become Y Combinator's fastest-ever unicorn at a $3.2B valuation, and watching that pace up close gave me the confidence to go build something of my own.
I helped drive fundraising for SMU's global stage where student founders from around the world pitch real business plans.
A parallel track through university. I basically did the full finance-internship rotation, corporate finance, investment banking, multi-strategy hedge fund exposure in my first year, equity and FX research at a boutique fund, plus a few case competitions, some consulting work, and a couple of selective bank programs along the way. Less about any single line on a resume, more about the people, sharp people from LSE, Columbia, Princeton, Wharton, and UCL. Honestly, there was something validating about holding my own next to people from schools like that, and it gave me a lot of confidence before I fully committed to building instead.
My first real startup rodeo. I joined as an Analyst and got promoted to Partner within four months. Content, UI/UX brainstorming, outreach, a bit of everything, learned on the job. Grateful to Johannes and Yuhana for scouting me on LinkedIn and giving me the shot.
My first ever internship, and a memorable one. It's also what told me, pretty clearly, that consultancy and audit weren't where I wanted to build a career.
Things still in the workshop, and things I've shipped.
Solith
Trust infrastructure for autonomous agents in crypto markets.
Crypto's AI agent ecosystem is running into three failure points. Fragmented data agents can't verify, stateless architectures that erase context between sessions, and no auditable trail for how an agent actually made a decision. None of that is a model problem. It's an infrastructure problem.
I'm building this for crypto-native institutions and builders who need agents that can be audited, that remember, and that operate on data they can actually trust.
Early days. Right now this is more thesis than product, and I'm still working out exactly how to build it properly. Putting it here anyway, because that's kind of the point.
CryptaVision
A risk-intelligence engine for digital asset investors. Real-time on-chain data, off-chain signals, actor behavior, and protocol dependencies, all pulled into one place instead of five different dashboards.
- Composite risk scores
- Narrative-driven explanations
- Interactive dependency maps
- Event-triggered early warnings
Before we wrote a line of code, we spent months just talking to people. 45+ trading desks, including Pantera, Delphi, and Amber, just to check the problem was actually real before building a company around it. Turned out it was, so we went ahead and bootstrapped it, with an ex-Flow Traders quant thinking through the markets side and a small team of ex-Goldman Sachs data engineers doing the unglamorous work of actually indexing the chain. Somewhere along the way we ended up on Grafa Crypto's Token2049 live broadcast, which was a pretty surreal place to be explaining risk scores to a room full of people. Didn't ultimately work out, but it's still the thing I learned the most from.
HAOO
A conviction-markets primitive, recognized as a top builder in the category within Outlier Ventures' ecosystem. Didn't take off, but it's where I learned the most about building ahead of a market that isn't ready for it yet.
View repo →claimcheck
An earlier experiment in the same problem Solith is now tackling head-on. Paste a claim, get back a trust verdict with evidence, a reasoning trace, and a tamper-evident audit trail.
View repo →Also shipping in the open
Events I show up to, and what I actually think about them.
More coming soon
Live on Grafa Crypto's Token2049 broadcast
Sat down on camera in Marina Bay to talk through CryptaVision and how we were thinking about on-chain risk. Still gets me every time I rewatch it, that's a lot of people to be explaining risk scores to.
Notes from building, mostly after the fact.
Building in AI or crypto? Let's talk.
Always up for a conversation about on-chain risk, agent infrastructure, or the next thing worth building. Reach out on whichever of these you actually check.