Applied Scientist

Location: Remote

Compensation: 2,000,000 9,000,000 per year

Description

Huzzle is building the next generation of human-data and agentic AI systems used by frontier AI labs and enterprises. We are looking for an Applied Scientist to design, train, and evaluate agentic systems that can reason, plan, and act across complex, real-world workflows.

In this role, you will work on reinforcement learning, preference optimization, and long-horizon agent behavior, while helping build realistic environments, datasets, and evaluation frameworks. Your work will directly shape how modern AI systems are trained, validated, and deployed in production settings.

Youll operate at the intersection of research and executionworking on ambiguous, high-impact problems and collaborating closely with engineers, researchers, and product leaders to turn ideas into working systems.

Key Responsibilities

• Design and build agentic systems for conversational and non-conversational workflows
• Implement advanced optimization techniques including supervised fine-tuning, instruction tuning, and preference optimization (e.g. DPO / IPO)
• Curate datasets and tools for model customization and post-training workflows
• Build evaluation pipelines for agent behavior, including automated benchmarks, multi-step reasoning tests, and safety checks
• Develop agentic architectures (e.g. CoT, ToT, ReAct) integrating planning, tool use, and long-horizon reasoning
• Prototype and iterate on multi-agent orchestration frameworks
• Collaborate with engineering and research teams to bring ideas into production
• Stay current with research in LLMs, RL, and agent-based systems, and translate it into practical systems

About the Team

Huzzle works with frontier AI labs and enterprises to produce high-quality human data, evaluation frameworks, and realistic agent environments. Our work focuses on long-horizon professional workflows, where trust, reliability, and correctness matter.

We are a small, fast-moving team. We believe data quality is the limiting factor for modern AIand we design our processes to let the data speak for itself.

Basic Qualifications

• Strong experience building applied ML or AI systems for real-world use cases
• Experience with Python, and at least one other systems or ML-adjacent language
• Hands-on experience designing experiments and analyzing results
• Familiarity with modern LLM training or post-training techniques
• Ability to work independently on open-ended, ambiguous problems

Preferred Qualifications

• Experience with reinforcement learning, preference optimization, or agentic systems
• Experience building evaluation or benchmarking pipelines
• Experience working on production ML systems
• Familiarity with multi-step reasoning, tool use, or long-horizon tasks

Why Huzzle

• Work on problems that directly shape how AI systems are trained and evaluated
• Collaborate with leading researchers, engineers, and AI labs
• High ownership, high impact, minimal bureaucracy
• Competitive compensation and rapid growth opportunities