Hiring nowPost-training and preference optimisation

Post-Training Research Engineer

Also called Research Engineer, Production Model Post-Training · Reward Model / Post-Training Engineer · RLHF Engineer · Fine-tuning Engineer

A Post-Training Research Engineer adapts pretrained models into ones that behave as the company needs. They turn preference data and reward signals into measurable gains in task quality without breaking safety or existing capability.

What they do

They take a pretrained model and shape its behaviour using demonstrations, preference data, reward models and reinforcement learning. Most days are spent designing experiments, launching training runs, reading results and working out why a change helped or hurt. They build reproducible comparisons so the team can trust what it sees. They debug unstable runs, tune hyperparameters, and check for reward hacking. They run regression and safety evaluations before anything ships. They write up what changed in the model's behaviour and why, so product and research colleagues can decide what to do next.

What a strong hire delivers

In the first months, a strong hire gets a reproducible post-training pipeline running end to end. They ship a post-trained model that improves on the target task against a clear baseline. They produce a behaviour change report that shows gains, regressions and safety results in plain terms. They also leave the team with trusted evaluation sets and experiment records, so later decisions are faster and better argued.

How you can tell they are good

Ask them to walk through one post-training project from baseline to result. Strong people explain how they controlled for noise, what they ablated, and what went wrong. They can describe a time a reward signal was gamed and how they spotted it. They separate real gains from evaluation artefacts. Worry if they quote only headline scores or cannot say what they would not trust.

Why it is hard to fill

Very few people have run preference optimisation and reinforcement learning on production-scale models, and most of them sit inside a handful of labs. Titles vary widely: RLHF engineer, fine-tuning engineer, reward model engineer, research engineer. Many candidates have fine-tuned small models but never handled distributed training or rigorous evaluation. Others are strong researchers who have never shipped. Expect a thin pool and long conversations.

Where to find them

Look in frontier labs and model startups, where post-training teams do this work daily. Adjacent pools include applied research teams fine-tuning open models, alignment and evaluation groups, and machine learning engineers who moved from recommender or ranking systems into reinforcement learning. Academic researchers in preference learning or reinforcement learning can also fit if they have built serious training infrastructure.

Indicative pay

Not enough published evidence yet to give a fair range for this role, which is common for work this new. We benchmark it for you as part of a search.

Interview questions for a Post-Training Research Engineer

  1. Describe a post-training project where your first approach failed. What did you change?

    A strong answer names the failure mode, such as unstable training, reward hacking or a weak data signal, and the evidence used to diagnose it. They explain the fix and what they learned. Worry about blame placed on data or compute alone.

  2. How do you make two post-training runs a fair comparison?

    Look for fixed seeds and data order where possible, repeated runs to estimate noise, matched compute, and held-out evaluations. They should track configs and code versions. Worry if they compare single runs or trust one metric.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We are hiring a Post-Training Research Engineer to shape how our models behave. You will take pretrained models and adapt them using demonstrations, preference data and reinforcement learning. The work is hands-on and evidence-led, and your results will decide what we ship.

What you will do

  • Run instruction tuning and preference optimisation experiments on our models
  • Build and maintain reward models and reinforcement learning training loops
  • Design reproducible comparisons and track every experiment
  • Run regression and safety evaluations before each release
  • Investigate unstable runs and reward hacking
  • Write clear behaviour change reports for research and product colleagues

What you will bring

  • Demonstrated work fine-tuning large models with supervised and preference-based methods
  • Strong PyTorch skills and experience with distributed training
  • Solid grounding in reinforcement learning and preference learning
  • Experience designing evaluations and separating real gains from noise
  • Clear written communication of experimental results

Nice to have

  • Experience with RLAIF or synthetic preference generation
  • Published or shared work on alignment or post-training
  • Experience building evaluation tooling

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Skills that matter

  • Reinforcement learningAdvancedOptimise policies from interaction and reward
  • Preference learningAdvancedLearn behaviour from comparisons and reward signals
  • Deep learningWorkingBuild and diagnose trainable neural models
  • Model evaluationWorkingMeasure capability and failure modes credibly

Other roles in post-training and preference optimisation

Hiring nowReward Modelling ResearcherA reward modelling researcher builds models that learn what good output looks like from preference comparisons. They give post-training teams a…Hiring nowReinforcement Learning Research EngineerA Reinforcement Learning Research Engineer improves how a model behaves by training it against interactive signals. They matter because reward gains…

Related roles in other families

EmergingReasoning Research ScientistReasoning, planning and searchA Reasoning Research Scientist tests and builds methods that make a model's multi-step inference reliable. They show the company which approaches…Hiring nowPlanning Algorithms ResearcherReasoning, planning and searchA Planning Algorithms Researcher builds planners that choose sequences of actions under hard constraints. Companies hire them so agents and systems…EmergingInference-Time Compute ResearcherReasoning, planning and searchAn Inference-Time Compute Researcher decides how much thinking a model should spend on each problem. They raise accuracy on hard tasks while keeping…EmergingNeuro-Symbolic AI ResearcherReasoning, planning and searchA Neuro-Symbolic AI Researcher combines learned models with symbolic rules and solvers, so systems reason consistently where rules matter. That…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert.

← Post-training and preference optimisationAll new job families

Start a search

Hiring a Post-Training Research Engineer?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.