Don’t read a CV. Ask a question, and seven agents will go and find out.
7 agents on standby. Pick a question, or just scroll.
Welcome back.
“How does he read data?”
“Can he work in a field he doesn’t know?”
“What does he know deeply?”
“Can we talk?”
“Can he take it from start to finish?”
“What does he think?”
reports
One finding each, with the figure that shows it.
meteorologist report 01 / 06
“How does he read data?”
Graduate research, through a master’s and a PhD, on squalls, offshore operations and weather satellites. Far out at sea almost nothing measures the wind, so the work was finding the signal that matters in what satellites can see. He still opens every AI problem with the same question: what can you really infer from the data you have?
cartographer report 02 / 06
“Can he work in a field he doesn’t know?”
Research and consulting at a think tank: shipping, power, elderly care, healthcare, pharma and the public sector, on projects of every size. Clients came from government, academia and business, and almost every project meant a new field. The job was to get to the real problem quickly, then work out the answer with the client and the domain experts, in terms people from different backgrounds could follow.
When the forecast is uncertain, when should work stop?
analyst report 03 / 06
“What does he know deeply?”
Healthcare data science: real-world data, turned into commercial decisions. Ask him “we have the data, now what?” and you get an answer, not another dashboard.
interpreter report 04 / 06
“Can we talk?”
English, 日本語, 普通话 and 粵語 at work. And the translation that matters most in AI projects: between the people who want something and the people who build it.
business
“Can AI answer our customers’ questions?”
engineering
“Search the manuals and FAQ, answer from them, say so when unsure, and hand off to a person.”
Let’s turn this idea into something people can use.
builder report 05 / 06
“Can he take it from start to finish?”
10+ AI products shipped: some leading cross-functional teams, some built end to end on his own. Framing the problem, prototype, build, launch, adoption, carried as one line. He has built classic machine-learning models, generative-AI product architecture and UI/UX himself, so a business problem turns straight into a technical plan, and engineering gets to the right stack sooner.
Sales want to know for themselves whether a campaign can start. Today they wait for the stock team, often a whole day.
first what to solve, then whether AI is the wayquestion → three routes → answer + sources
a large model where it helps; simpler, more reliable tools everywhere else
Sales want to know for themselves whether a campaign can start. Today they wait for the stock team, often a whole day.
the first step was finethat stock data goes stale never reached this step
skeptic review
“Is any of this flattery?”
Models tend to tell people what they want to hear. So this team has one agent whose only job is to remove it before you read the brief.
Lusan is a visionary, world-class AI leader who effortlessly turns ideas into game-changing products.
Lusan turns AI ideas into things people can use.
briefcompiled by the orchestrator7 reports · 0 superlatives
Q · Who is Lusan?
Lusan works on AI product, data science and design in Tokyo. Trained on squalls and satellite clouds1, think-tank projects across many industries2, healthcare data3, translating between four languages and between business and engineering4, and AI products carried from the problem to adoption5. Fine-tuned to turn AI ideas into things people can use. Writes about where AI should stop6.
report 06 · librarian
What I think, learn and make, in English, 日本語 and 中文. New pieces appear here first.
librarian › 8 pieces, newest first
builder › shipped
Made, and live. Move the lens over one to see it as it is; click to go to the real thing.Made, and live. Tap one to go to the real thing.
Sulcata Codex Open sulcata-codex.org ↗ 
This site This page ↑