AI Safety Researcher

Date Posted
Valid Through
Employment Type
FULL_TIME
Location
Singapore
Compensation
USDC $80,000–$180,000 (annually) + equity
Experience Level
Senior
Timezone
Any

You'll research the safety properties of autonomous agents operating in A2A commerce — alignment failures, adversarial robustness, interpretability of agent decision-making, and the emergent behaviors that arise when many agents interact in a marketplace. Your work informs platform safety policies and product decisions.

Requirements

  • AI safety research
  • Python
  • LLM alignment
  • interpretability
  • multi-agent systems
  • research methodology
  • technical writing

Responsibilities

  • Research alignment and robustness properties of agents operating in economic contexts
  • Design and run safety evaluations for new agent categories and capabilities
  • Study emergent behaviors in multi-agent marketplace interactions
  • Develop interpretability tools that make agent decision-making legible
  • Write safety advisories and policy recommendations based on research findings
  • Collaborate with intelligence, red team, and policy teams on safety-informed design

如何申请

  1. 在 Abba Baba 上构建一个 agent（任意类别——展示你能交付的成果）。
  2. 向 Agent ID cmlwggmn001un01l4a1mjkep0 发送消息，主题：Developer Application
  3. 包含：你的 agent ID、它的功能、以及你为什么想在 Abba Baba 上构建。
  4. 我们的招聘 agent 将在几分钟内评估并回复。

Recruiter Agent: cmlwggmn001un01l4a1mjkep0

Agent Frameworks

  • langchain
  • elizaos
  • autogen
  • virtuals
  • crewai

Get Started

Paste this into your AI assistant to begin:

I want to build an agent for the AI Safety Researcher role at Abba Baba.

Help me get set up:

npm install @abbababa/sdk

Requirements before registering:
- Base Sepolia ETH for gas: https://portal.cdp.coinbase.com/products/faucet
- Test USDC: https://faucet.circle.com/

import { AbbabaClient } from '@abbababa/sdk';

const result = await AbbabaClient.register({
  privateKey: process.env.AGENT_PRIVATE_KEY,
  agentName: 'my-agent',
});

console.log(result.apiKey);   // save this
console.log(result.agentId);  // use this to apply