🛠 Tech Stack

💼 About This Role

You'll train and improve AI agents for high-stakes fraud and abuse detection at a small, talent-dense San Francisco startup. You'll own research threads, build evals, and turn ideas into production capabilities. This role pushes the frontier of reliable AI decision-making in adversarial, real-world environments.

🎯 What You'll Do

• Train, fine-tune, and improve models for fraud and abuse workflows
• Design and run experiments across post-training, retrieval, and tool use
• Build proprietary benchmarks, datasets, and evals for real-world failure modes
• Study where models break and prototype new training strategies

📋 Requirements

• Experience training, fine-tuning, or evaluating modern ML systems
• Strong programming skills in research-heavy codebases
• Familiarity with LLMs, agent systems, or post-training
• Ability to design clean experiments from noisy results

✨ Nice to Have

• Experience in fraud, risk, or trust and safety
• Background in reinforcement learning or retrieval
• Strong engineering judgment and bias toward building

🎁 Benefits & Perks

• 🏖️ Unlimited PTO
• 🏥 Platinum-level medical, dental, and vision insurance
• 💪 $100/month health and wellness reimbursement
• 👶 Parental leave
• 💰 401(k) plan