Applied ML Research Intern - Post-Training & RL Experiments

NextToken AI Inc.; San Francisco, Northern (CA, KY); Ibrido; USD 41.000 - 83.000; Tempo pieno

Descrizione del lavoro

NextToken AI Inc. seeks a motivated student or early graduate in CS/ML to contribute to evaluation-heavy ML research. You will design experiments using real production traces, build verifier and reward models, and run end-to-end post-training workflows with SFT, RL, and ablations.

The role emphasizes rigorous evaluation, bias-aware benchmarking, and a bias-to-ship mindset, aiming to translate research into deployed models or concrete results.

Competenze

• Strong empirical ML background with hands-on experiments on LLMs or related models.
• Ability to design and justify evaluation benchmarks that avoid biases.
• Comfort working with real production traces, not curated datasets.

Mansioni

• Build evaluation suites from production traces and define metrics.
• Develop verifiers and reward models to guide training.
• Run end-to-end post-training experiments (SFT/DPO/RL), ablations, and analysis.

Conoscenze

Empirical ML

RL experiments

Evaluation benchmarks

Production data handling

Diagnostic reasoning

Formazione

BS/MS/PhD in CS/ML

Strumenti

RL environments

Verifier models

Descrizione del lavoro

NextToken AI Inc. seeks a motivated student or early graduate in CS/ML to contribute to evaluation-heavy ML research. You will design experiments using real production traces, build verifier and reward models, and run end-to-end post-training workflows with SFT, RL, and ablations.

The role emphasizes rigorous evaluation, bias-aware benchmarking, and a bias-to-ship mindset, aiming to translate research into deployed models or concrete results.