NextToken

Post-training systems for continuous AI improvement

Applied Research Intern

Applied Machine LearningConfirmed live in the last 24 hours

InternshipNo salary listedIn-person

About the Job

Requirements

Currently be a student pursuing a Bachelor of Science, Master of Science, or PhD in Computer Science, Machine Learning, or an adjacent field, or have recently graduated.
Have a strong empirical machine learning background and experience running post-training or reinforcement learning experiments on large language models, with the ability to explain the results.
Be rigorous about evaluation and able to identify flawed benchmarks and design reliable evaluations.
Be comfortable using production data traces rather than curated datasets.
Be available for at least 12 weeks at meaningful intensity.

Responsibilities

Build evaluation suites from production traces, decide what to measure, build graders, and demonstrate that the graders track real quality.
Build verifiers and reward models by converting human corrections, outcomes, and standards of good work into training signals.
Run end-to-end post-training experiments, including forming hypotheses, preparing data, running supervised fine-tuning, direct preference optimization, and reinforcement learning, performing ablations and analysis, and writing up results with quantitative evidence.
Build reinforcement learning environments and agentic harnesses for workflows targeted by specialized models.
Reproduce recent post-training, reasoning, and agent research papers, and integrate successful approaches into the company's stack.
Drive experiments toward a deployed model or a decision to stop pursuing a hypothesis.

Desired Qualifications

Have publications or open-source contributions in post-training, reinforcement learning, evaluations, or agents.
Have experience with verifier models, synthetic data, or reinforcement learning environments at scale.
Have previously shipped research into a production system.
Have strong judgment about which areas the field is over-investing in or under-investing in.

Sit at the edge of research and production. You'll take a real workflow from a real team, make its failure modes measurable, construct the reward system, and post-train models that beat the frontier baseline on the work that matters. We hire interns who want to be researchers who ship.

What you'll own

Evaluation suites built from production traces: deciding what to measure, building the graders, and proving they track real quality.
Verifiers and reward models. Turning human corrections, outcomes, and standards of good work into training signal.
Post-training experiments end-to-end: hypothesis, data, SFT/DPO/RL runs, ablations, analysis, and the write-up that defends the win with numbers.
RL environments and agentic harnesses for the workflows we're specializing models for.
Reproducing recent post-training, reasoning, and agent papers, and translating the ones that work into our stack.

About you

Currently a student (BS / MS / PhD) in CS, ML, or an adjacent field, or recently graduated.
Strong empirical ML background. You've run post-training or RL experiments on LLMs and can explain what worked, what didn't, and why.
Rigorous about evaluation. You can spot a benchmark that lies and design one that doesn't.
Comfortable in messy production data. Real traces, not curated datasets, are your raw material.
Bias to ship. Your experiments end in a deployed model or a killed hypothesis, not just a doc.
Available for at least 12 weeks at meaningful intensity.

Bonus

Publications or OSS in post-training, RL, evals, or agents.
Experience with verifier models, synthetic data, or RL environments at scale.
Shipped research into a production system before.
Strong taste for what the field is over- vs. under-investing in.

About the Company

NextToken

Post-training systems for continuous AI improvement

NextToken helps AI product teams continuously improve models after deployment by creating a feedback loop from production to the model, turning general-purpose AI into specialized high-performance agents. Its platform converts production traces, expert feedback, and real-world tasks into structured evaluations and training data, then applies post-training methods like supervised fine-tuning, RL with verifiable rewards, policy optimization, and model distillation. It differentiates itself with an end-to-end workflow that ties production feedback directly to model updates, enabling organization-specific intelligence and practical deployment of optimized agents. Goal: help teams deploy and steadily improve specialized AI models with high accuracy in real-world tasks.

Company size: 1-10
Company stage: Not listed
Total funding: Not listed
Headquarters: Austin, TX
Founded: 2016

Recently Posted Jobs

Explore more opportunities at NextToken

See more jobs