A lot has to happen between a caller finishing a sentence and the agent starting one. Our code detects the end of speech, transcribes it, lets a model choose a tool and synthesises the answer; theirs turns that tool call into a write to their booking system, all inside a latency budget short enough to feel like conversation. I build across both: the voice pipeline, the agent architecture, the integrations, and the evals that catch regressions before callers do.
This caller only wanted to move a booking. Behind the scenes everything worked, but the agent missed two short replies and they ended up asking for a person. A lot of my client work happens at this level: going through calls with their support team, agreeing what a good call sounds like, and fixing what we find until the agent handles calls the way their best people do.
I usually meet a customer’s support team weeks before the agent answers a call. I work on the whole account, not just the build: discovery calls to learn how they work, scoping which calls the agent takes first, the plan we agree on, weekly syncs through launch, and expansion, which starts when they ask whether it could take cancellations too.
I’m a forward deployed engineer at Giga, here in San Francisco. Giga builds AI agents that answer customer support calls for big companies, and I joined as employee number 14. The streets don’t line up across Market Street because in 1847 the landowners on each side wanted their own grid, and both got one.
The best thing about the Bay is who turns up. I found it easy to start a monthly meetup for the Bay’s energy founders, investors and researchers: the Energy Unconference, where whoever turns up sets the agenda. Later I led strategy and operations for The AI Collective. I also interned at BART, building a model that warns before a control room overheats.
I came to California in 2021 to study political economy at Berkeley. Everyone around me seemed to be building something, so I switched to computer science, and I’ve stayed for the same reason. It also helps that Tahoe is a few hours away. Back to the world
I was born in Hong Kong, went to school in the UK and now live in California, where I build enterprise agents at Giga. I studied computer science at Berkeley and speak English, Cantonese and Mandarin. Outside work I read about trade and build software for it.
Zoom out far enough and the world is plumbing. Oil funnels through the Strait of Hormuz, grain queues for the Bosporus, and copper leaves Chile by sea. On the side I build small tools for the paperwork behind all of it: laytime clauses, compliance bulletins, tariff codes. It still runs on mostly PDFs.
Markets are where history is written in numbers. The institutions that keep those books, from trading houses to insurers and ports, still run on email and spreadsheets. I think they’ll be rebuilt as software. If you’re thinking about this too, I’d like to compare notes.
I was born and grew up here. The minibuses won’t stop unless you shout, lunch at a cha chaan teng often means sharing a table with strangers, and a beach or a ridge trail is never more than an MTR ride away. The city assumes you know where you’re going. Back to the world
I went to school in the UK, first at Rugby, where the game was invented, then at Sevenoaks, half an hour from London. I played rugby, came second in a pub quiz, and spent many summers watching England almost win something. I also got curious about economics there, and wrote essays on questions like whether cities have overtaken countries in the world economy. One was shortlisted for the NCH essay prize in economics. Back to the world