Applied AI Research Engineer: RL & Data Pipelines

Sonder; New York (NY); On-site; USD 250,000 - 300,000; Full time

Benefits offered by this job

Health insurance

Dental insurance

Vision insurance

Flexible PTO

Relocation assistance

Job summary

Sonder, an applied AI lab in New York, seeks a Research Engineer to push model capabilities via data, training environments, and RL experiments. You will own experiments from hypothesis to measurable improvements, collaborating with researchers and engineers to close the loop and build reliable tooling.

We value Python engineering strength, RL know-how, and hands-on experience with data generation, synthetic data, and interactive environments. In-person collaboration in NYC is expected.

Qualifications

• Strong Python engineering skills and hands-on experience training or post-training language or multimodal models with PyTorch, JAX, or a comparable framework.

Responsibilities

• Build high-quality data pipelines for generation, curation, and validation of training data.
• Develop reliable training environments for multi-step tasks with reproducible execution and feedback.
• Improve post-training with supervised baselines, RL experiments, and analysis of data, rewards, and optimization choices.
• Design graders and held-out evaluations to ensure trustworthy measurements and detect reward hacking or data contamination.
• Close the experimental loop by inspecting model failures, forming testable hypotheses, and building reproducible tooling.

Skills

Python programming

Reinforcement learning

Experimentation & evaluation

Data pipelines

Tools

PyTorch

JAX

Job description

Sonder, an applied AI lab in New York, seeks a Research Engineer to push model capabilities via data, training environments, and RL experiments. You will own experiments from hypothesis to measurable improvements, collaborating with researchers and engineers to close the loop and build reliable tooling.

We value Python engineering strength, RL know-how, and hands-on experience with data generation, synthetic data, and interactive environments. In-person collaboration in NYC is expected.