Contact ↗

ALLEN
ZHOU

Jarvis AI agent product screen

AI-native founder exploring new patterns for human-computer interaction.

◎ Global-minded builder
Mandarin + English
Exploring AI-native interfaces

Shaped by finance, ML, and product systems across cultures
A few artifacts from the current build.

Selected
Works

Experience

10+ years across AI-native product, recommendation systems, growth ML, financial data, and investment workflows.

Jarvis logo AutoApe AI logo
Ant Group logo
Uber logo
KPMG logo
Deutsche Bank logo
001 Jarvis
Founder / AI Product Lead

Built an agentic finance application with streaming tool calls, typed tool definitions, per-tool result UI, wallet/auth flows, transaction simulation, and visible execution state.

002 AutoApe AI
Founder / Product Builder

Built and demoed browser-task automation flows that turn natural-language intent into visible web actions, product scripts, evaluation examples, and user-facing automation narratives.

003 Ant Group
Algorithm Expert

Deployed merchant classification across 100M merchants, expanded coverage by 20M at 0.76 F1, led five engineers on Double-12 recommendation systems, and drove 10M incremental users.

004 Uber
Growth Data & Analytics

Built RandomForest and XGBoost cross-sell models over roughly 50M behavior records, reached 0.81 AUC, improved campaign conversion by 40%, and coordinated Singapore/SF rider-app measurement.

005 KPMG
M&A / Valuation

Built transaction models and investor materials for $25M-$125M deals, with API-fed DCF tools, scenario analysis, and commodity-price regression to connect narrative with measurable drivers.

006 Deutsche Bank / RL
Financial Systems

Shipped Java and VBA workflow tools for derivatives regulatory reporting and analytics, cut repeated reporting work by roughly 90%, and automated investment research refreshes with APIs and Bloomberg inputs.

Latest
Notes

A short preview of writing on agent UX, product education, and AI-native interaction.

Latest notes
001Tool-calling UXTrustworthy agent interfaces start before the tool runs: intent, typed inputs, progress, evidence, failure, and control. 002Chinese product noteComplex AI systems become useful when builders can follow a reproducible path, not just a translated concept.
View all writing

Approach [How I work]

I focus on ambiguous technology that needs to become legible, useful, and worth trusting.

001AI-native product systemsBuilding interfaces where models, tools, memory, state, and user control work as one product surface.
002Agent interaction designMaking intent, progress, evidence, and action boundaries visible so powerful systems feel inspectable.
003ML and growth systemsTurning noisy behavioral data into classification, recommendation, ranking, measurement, and lifecycle systems that can survive real operating pressure.
004Cross-cultural product senseMoving between Chinese and English contexts without flattening the product into translation or slogans.
I'm Allen Zhou

I'm an AI-native founder and ML engineering leader working across product development, machine learning, finance, and new interaction models.

Build the artifact. Explain the path.

Contact Resume GitHub @agentprompt