Machine Learning Researcher (Agentic Post-Training for On-Device Models)

Quick Facts

• Role: Machine Learning Researcher (Agentic Post-Training for On-Device Models)

Description

You will develop post-training methods that turn foundation models into capable, reliable, and efficient agents for on-device execution. The work includes supervised fine-tuning and reinforcement learning methods, plus the training data, interactive environments, reward signals, and evaluations needed to improve reasoning, tool use, and long-horizon tasks under tight latency, memory, and energy budgets.

Responsibilities

• Advance agentic post-training for on-device models.

• Develop supervised fine-tuning, reinforcement learning, preference optimization, and distillation methods to improve reasoning and multi-step tool use in on-device LLMs.

• Design training data and interactive environments (trajectory generation, curriculum construction, hard-example mining; multimodal and tool-use learning environments).

• Design reward functions, verifiers, and graders for complex agent tasks, including long-horizon credit assignment and alignment between reward and real task success.

• Improve capability under deployment constraints (balance task success with latency, memory, energy, and tool-use cost; use structured/compressed reasoning and adaptive computation).

• Build rigorous evaluations and diagnostics (measure success, generalization, reliability, and efficiency; analyze trajectories to isolate issues and drive improvements).

• Translate research into on-device experiences through collaboration with software and hardware teams.

• Drive external research impact by publishing high-impact technical reports and releasing open-source models.

Requirements

• MS in Computer Science, Machine Learning, Artificial Intelligence, Computer Vision, Natural Language Processing, or related field (or equivalent practical experience).

• Strong research track record: first-author publications at top-tier conferences (NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, ACL, EMNLP) or substantial hands-on experience training frontier models, with relevant contributions to foundation models, reinforcement learning, multimodal learning, or agentic systems.

• Strong Python programming skills and proficiency with PyTorch.

• Demonstrated ability to formulate hypotheses, implement training and evaluation pipelines, conduct controlled experiments, and draw well-supported conclusions.

Benefits

• End-to-end ownership of research projects: identify capability gaps, build training pipelines, run rigorous experiments, and translate results into product value.

• Opportunities to publish technical reports as a core author and release open-source models that support the broader research community.