Research, Coding Agents
Thinkingmachines👥 51 - 200 employees🏢 Technology
Join a high-leverage team to enhance AI coding capabilities through research, design, and execution of RL training jobs and data generation.
Research, General Agents
🕒 8 days ago
PythonDeep Learning FrameworksSynthetic Data PipelinesModel Usability
📜 Description
- Build a synthetic data framework for training agentic capabilities.
- Create new task environments to enhance tool use and complex workflows.
- Address gaps through data and recipe work, validating changes with controlled ablations.
- Translate feedback into model improvements for enhanced usability.
🛠️ Requirements
- Strong engineering skills with the ability to debug complex codebases.
- Experience designing, running, and interpreting experiments scientifically.
- Proficiency in Python and familiarity with deep learning frameworks like PyTorch or TensorFlow.
- Bachelor's degree or equivalent experience in Computer Science, Machine Learning, or related fields.
- Clear communication skills to explain complex technical concepts.
✨ Benefits
- Dental
- Vision benefits
- Unlimited PTO
- Paid parental leave
- Relocation support as needed
Similar jobs
Search more Research Scientist jobsResearch, Post-Training Evals
Thinkingmachines👥 51 - 200 employees🏢 Technology
Develop reliable model evaluations for research signals, focusing on evaluation creation, usability, auditing, and efficiency in collaboration with researchers and engineers.
This role focuses on scaling reinforcement learning for frontier models, requiring expertise in asynchronous RL algorithms and the integration of training and inference systems.
Research, Pre-Training Data
Thinkingmachines👥 51 - 200 employees🏢 Technology
Pre-training researchers will design and implement methods for sourcing, curating, and analyzing pre-training datasets, blending research with large-scale data engineering.
As a safety researcher, you will ensure the safety and trustworthiness of AI models by designing experiments and developing techniques to handle harmful requests effectively.