AI Team Org Map — Kaon (April 2026)

Maintained as living doc. Last updated: 2026-04-06

All teams report to Alex Xi (CTO) Total AI headcount: ~28 active


1. 后训练 / Post-Training

Lead: 韩子扬 ICs: 陈昱, 李家宜, 王钦语, 陈佳新, 刘俊文, 郝少春, 1 intern Headcount: 7 ICs + lead Focus: SFT, GRPO, reward modeling, training data pipeline

2. 个性化 / Personalization

Lead: 韩子扬 + 杨千驰 (co-lead) ICs: 刘左茜, 付泓鑫, 新人 Headcount: 3 ICs + co-leads (韩子扬 dual-hatting with Post-Training) Focus: User embedding, personalization, population intelligence

3. AI工程 - Agentic

Lead: Chen Xi ICs: 李大山, Xinyi Chen, 叶沛津, Leon, 田勇, 谢靖 Shared: 付泓鑫, 刘左茜 (shared with Personalization) Headcount: 6 ICs + lead (+ 2 shared) Focus: Agentic runtime, agent systems, Memory engineering, 个性化 engineering, tool call pipeline Note: Owns the full agent stack end-to-end — runtime, memory, personalization, all capability skills

4. AI工程 + 平台 / AI Platform

Lead: 李斌 ICs: 范晓豪, 李铭宇 Headcount: 2 ICs + lead Focus: Internal platforms — n8n, Shiki, Evals, Wagashi, Tianji

5. AI Infra

Lead: 田浩廷 ICs: 王谦 (inference optimization) Away: 张越 (intern, inference optimization — temporarily away) Headcount: 1 IC + lead (+ 1 away intern) Focus: Conv AB, ELO, LSP, GPU management, data infra, inference optimization, agent infra

6. 多模态 / Multimodal

Lead: Boyn Chan + Felix (co-lead) ICs: 伍袁锋 (Curry), 刘欣然 Headcount: 2 ICs + co-leads Focus: Image generation, multimodal capabilities

7. AI Product

Person: 牟文钊 Working with: Water Focus: AI product direction

8. RM / Eval ⚠️ GAP

Lead: None — needs hire Team: No dedicated team yet (partially handled within Post-Training) Target Focus: Reward model training, eval benchmarks, A/B test design, bad case triage, quality signal definition


Key Gaps & Hiring Needs

PriorityRoleTeamLocationEst. CostNotes
🔴 P0RM/Eval LeadNew team (#8)US or ChinaT2 ($250-350K TC) if USDefines “what is good” — critical for all model work
🟡 P1AI Infra engineerAI Infra (#5)China~$8K/moBackfill for 何川 departure; 田浩廷 nearly solo
🟡 P1Personalization researcher个性化 (#2)China~$8K/moTeam is thin (3 ICs) for ambitious scope
🟢 P2AI engineers (1-2)Agentic (#3)China or USVariesAs KAI ramps, Chen Xi’s team carries heavy load
🟢 P2AI运营 / Data OpsCross-teamChina~$5-6K/moTraining data quality, annotation, prompt writing — currently scattered

Budget Context

  • Extra hiring budget: 300K/mo (on top of existing payroll)
  • China cost: ~$8K/mo per person
  • US cost: ~$25K/mo per person
  • 1 US hire ≈ 3 China hires