Machine Learning Researcher - Multi-Agent Systems
🌉 San Francisco, CA (On-site)
About Us
Softmax develops multi-agent reinforcement learning simulations focused on social learning scenarios. Our work explores how multiple agents can maintain individual identity while achieving coherence as a unified system—similar to multicellularity in nature.
The Role
We’re seeking a Machine Learning Researcher to design experiments, tune models, and analyze results within our multi-agent reinforcement learning framework. You’ll work closely with our engineering team to co-design infrastructure while focusing on the ML aspects of our research.
Key Responsibilities
- Design and execute experiments in multi-agent reinforcement learning
- Select and customize neural network architectures for specific learning tasks
- Tune models and hyperparameters to optimize agent performance
- Analyze experimental results and identify patterns in agent behavior
- Collaborate with engineers to ensure infrastructure meets research needs
- Translate research findings into actionable insights
Requirements
- Strong background in machine learning, particularly reinforcement learning
- Experience designing and running ML experiments
- Proficiency in at least one deep learning framework (PyTorch, TensorFlow, etc.)
- Ability to analyze complex data and draw meaningful conclusions
- Collaborative mindset and strong communication skills
Nice to Have
- Publications or significant projects in multi-agent systems
- Experience with emergent behavior or social learning models
- Familiarity with distributed training systems
- Background in complex systems or cognitive science
Company Culture
Our team values coherence, both in our technical work and our approach to collaboration. Many team members have backgrounds in practices that cultivate insight (meditation, contemplative practices, etc.), though this is not a requirement. We appreciate candidates who are thoughtful about systems — both technical and social.
Compensation
Salary between $150,000 to $300,000 per year, plus equity compensation.
