Member of Technical Staff, Model Training; applied research

In this role, you will work across language models, multimodal systems, robotics, and autonomous agents, developing models optimized for specific applications and environments. You will own the full model development lifecycle, from understanding application requirements and designing training datasets to model adaptation, evaluation, and production deployment. Our objective is to build specialized models that match or exceed the quality of frontier models on targeted workloads while significantly improving inference efficiency, latency, and cost.

Train, fine-tune, and adapt language, vision-language, and robotics models using supervised fine-tuning (SFT), reinforcement learning (RL), imitation learning, and distillation. Develop specialized models for complex applications, including multimodal understanding, robotic perception and control, and autonomous agents capable of reasoning, planning, and tool use. Build simulation-based training and evaluation environments to study policy learning, task generalization, and the transfer of capabilities from simulated to real-world settings.

Combine expert demonstrations, human feedback, synthetic datasets, and environment-generated trajectories into effective training datasets, reward signals, and learning curricula. Develop specialized models that power micro1's flow ecosystem, improving data generation, annotation, quality assurance, and automated grading workflows. Incorporate expert corrections and model performance feedback into iterative training cycles to continuously improve accuracy, robustness, and task success. Optimize models for production inference, balancing quality, latency, throughput, and computational efficiency against frontier-model baselines.

Collaborate with research and platform engineering teams to transition promising prototypes into scalable, production-ready model deployments.

• Model Training & Adaptation
• Multimodal AI
• Robotics & Agent Learning
• Applied ML Evaluation
• Inference Optimization & Model Deployment