Human Card — Katelyn Lv

What I Do

Map and recruit top AI researchers & engineers across World Model, Embodied AI, Speech Foundation Models, AIGC, and AI Infra

Focus Areas

World Model Robotics Deep Learning Brain-Computer Interface Embodied AI Data Infrastructure Product Strategy

Self-Built Systems

2000Candidates Tracked

29Benchmark Teams

24World Models Mapped

15+Knowledge Entries

Projects

🧠 Situated Memory — Domain-Based Agent Memory Architecture

Novel organizing principle for personal AI agent memory: life domains instead of cognitive mechanisms. 77% context reduction in work sessions. Position paper targeting CHI / CSCW.

Agent MemoryHCIPersonal AI

📊 Money Tracker — Embodied AI Talent Intelligence

Full-stack dashboard (Python + SQLite + vanilla JS) for tracking benchmark teams, world model groups, and professor labs. Collapsible domain tags, layered card views, search-to-expand navigation.

DashboardTalent IntelligenceSQLite

🛰️ Tech Radar — Market Signal Scanner

Automated pipeline monitoring startup funding, paper publications, and talent movements in embodied AI. Integrates with Feishu Bitable for collaborative signal processing.

Market IntelligenceAutomationFeishu

📝 Smart Notebook — CRM & Knowledge Base

Local-first knowledge management for candidate profiles, company deep-dives, and structured interview notes. Direct JSON manipulation for data integrity, bypassing the Electron GUI when needed.

CRMKnowledge ManagementElectron

AI Memory & Agent Research

Situated Memory

A research project exploring how AI agents should manage memory across sessions and contexts. Core thesis: memory is not a retrieval problem — it's a situation awareness problem. The agent must know what context it's in before it can decide what's relevant. Paper in progress (multi-user validation as key experimental design).

How I Work

Research-first: study the tech landscape (conferences, benchmarks, papers) before initiating outreach
Three-layer filtering: Academic track record → Industry impact → Domain-role fit
Structured deliverables: comparative analysis over vague recommendations
Tool builder mindset: if a system doesn't exist, I build it

Actively Looking For

Someone who can design and lead a public embodied AI benchmark / evaluation framework
Senior researchers in World Model, video-to-action, and physical simulation
Model-efficiency researchers (architecture-level, not infra-level) for large-scale training

Recent Updates

Jul 2026Launched Situated Memory — domain-based memory architecture for personal AI agents. Paper draft in progress.

Jun 2026Built Money Tracker — SQLite-powered talent & team tracking dashboard for embodied AI recruitment.

May 2026Designed Tech Radar — automated market signal scanner for startup funding & talent discovery.

2025–2026Developed AI-powered recruitment intelligence tools: smart notebook, company profiling, CRM automation.