tannah.geng

I'm Tannah GengTAI, AI product manager and builder in Paris.

I turn what AI can do into products people actually use.

Tannah's memoji
Multi-agent systems✦MCP✦LLM evals✦Human handoff✦RAG✦Voice AI✦Data products✦ML forecasting✦Prototyping with Claude Code✦0 → 1✦Multi-agent systems✦MCP✦LLM evals✦Human handoff✦RAG✦Voice AI✦Data products✦ML forecasting✦Prototyping with Claude Code✦0 → 1✦
10 years
in product, across 4 countries
Multi-agent
AI assistant in enterprise pilot
1 MVP
built end-to-end with AI coding tools
2,000+
monthly users of a data product in 6 countries

Selected work

( 04 )
01 Concept case study

Designing an agentic order-issue assistant for e-commerce

A full product case for a fictional marketplace, covering agent architecture, routing, evals and release gates. The core idea is to measure true resolution, not containment.

4 + 1specialist agents plus one human handoff, gated by evals
Open case study ↗
Shopper→Orchestrator
OrderReturnsSellerHuman
02 Amex GBT · 2025 – now

An AI traveler assistant for enterprise travel

I'm leading a multi-agent pilot at Amex GBT, with specialist agents, live trip data through MCP, and LLM-as-judge release gates.

Livetrip data through MCP, in enterprise pilot
Open case study ↗
PAR✈JFK
Tue 09:40 · Seat 14A · Gate K32
Answered from live trip data · MCP
03 Founder · 2023 – 2025

Building an AI assistant for ADHD, solo

I took it from problem to a working text-and-voice MVP on Telegram, built with Claude Code and Cursor.

1 MVPbuilt solo, by text and voice
Open case study ↗
Morning! You wanted to call the dentist today. Shall I remind you at 10?
▶0:04
Done. I'll call you at 10.
04 Lazada · 2021 – 2023

A logistics data product used in six countries

At Lazada, I raised operational productivity by 20%, launched ML volume forecasting and improved NPS by 20%.

+20%operational productivity across 6 countries
Open case study ↗
productivity

How I think

( 03 )
01Evals are the specFor AI features, the eval set is the most precise requirements doc you'll write.Read ↗
02The handoff is the productAn agent's most important decision is when not to answer.Read ↗
03Build to learn, not to launchWhat building an AI MVP solo taught me about scoping agents.Read ↗

Where I've been

Singapore → New York → Paris
  1. 2025 – nowAI Product ManagerAmex Global Business TravelParis
  2. 2023 – 2025Founder, AI assistant for ADHDHook / Antler incubatorsParis
  3. 2021 – 2023Senior Data Product ManagerLazadaSingapore
  4. 2019 – 2021Associate Product ManagerPostlightNew York
  5. 2018 – 2019MBACornell UniversityNew York
  6. 2014 – 2018Product Lead / Management TraineePacteraSingapore
  7. 2009 – 2013Bachelor of EngineeringNanyang Technological UniversitySingapore

Building something with AI? Let's talk.

I'm looking for AI product roles on agentic platforms and AI workflows, in Paris or remote.

tannah.geng@gmail.comLinkedIn ↗