Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.
A core focus of ours is agents that can learn to find their own objectives in the world. We are looking for researchers to design and build new ways of using RL where the formulation of rewards and tasks need to be discovered, rather than given.
Develop RL methods for agents that can discover useful objectives, tasks and curricula without relying entirely on human-specified rewards.
Design systems for open-ended learning, including unsupervised/automated environment design, asymmetric self-play, and intrinsic motivation.
Build training loops where agents learn from interaction, exploration, novelty, competence progress, self-generated challenges, or other nonstandard reward signals.
Investigate how agents can avoid collapse into trivial, degenerate, or easily exploitable objectives.
Own and develop a research agenda within Vmax, from identifying promising directions to executing experiments and communicating results.
PhD or equivalent experience in machine learning, reinforcement learning, artificial intelligence, or a closely related field.
Track record of strong technical work, demonstrated through publications, open-source projects, deployed systems, competitions, or equivalent contributions.
Deep understanding of reinforcement learning
Strong interest in open-ended learning
Experience with LLM post-training
Strong empirical research ability, including designing experiments, choosing meaningful baselines, running ablations, and diagnosing unexpected results.
Strong programming ability in Python and experience with at least one major ML framework such as PyTorch or JAX.
Ability to work independently on ambiguous research problems and turn high-level ideas into concrete experimental programs.
Ability to collaborate effectively with researchers and engineers on ambiguous, fast-moving technical problems.
Clear written and verbal communication of technical ideas, results, tradeoffs, and risks.
Experience with open-ended learning, automatic curriculum generation, intrinsic motivation, self-play, goal-conditioned RL, unsupervised skill discovery, multi-agent RL, quality-diversity, or evolutionary methods.
Familiarity with methods such as POET, population-based training, or multi-agent RL
Experience designing benchmarks or evals for generalization, exploration, long-horizon learning or behavioral diversity
Demonstrated taste for identifying non-obvious research directions and converting them into tractable experiments.
this role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement
The expected salary range for this position is $300,000 - $500,000 USD
","descriptionPlain":"ABOUT VMAX\n\nVmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.\n\n\nABOUT THE ROLE\n\nA core focus of ours is agents that can learn to find their own objectives in the world. We are looking for researchers to design and build new ways of using RL where the formulation of rewards and tasks need to be discovered, rather than given. \n\n\nRESPONSIBILITIES\n\n - Develop RL methods for agents that can discover useful objectives, tasks and curricula without relying entirely on human-specified rewards.\n\n - Design systems for open-ended learning, including unsupervised/automated environment design, asymmetric self-play, and intrinsic motivation.\n\n - Build training loops where agents learn from interaction, exploration, novelty, competence progress, self-generated challenges, or other nonstandard reward signals.\n\n - Investigate how agents can avoid collapse into trivial, degenerate, or easily exploitable objectives.\n\n - Own and develop a research agenda within Vmax, from identifying promising directions to executing experiments and communicating results.\n\n\nMINIMUM REQUIREMENTS\n\n - PhD or equivalent experience in machine learning, reinforcement learning, artificial intelligence, or a closely related field.\n\n - Track record of strong technical work, demonstrated through publications, open-source projects, deployed systems, competitions, or equivalent contributions.\n\n - Deep understanding of reinforcement learning\n\n - Strong interest in open-ended learning\n\n - Experience with LLM post-training\n\n - Strong empirical research ability, including designing experiments, choosing meaningful baselines, running ablations, and diagnosing unexpected results.\n\n - Strong programming ability in Python and experience with at least one major ML framework such as PyTorch or JAX.\n\n - Ability to work independently on ambiguous research problems and turn high-level ideas into concrete experimental programs.\n\n - Ability to collaborate effectively with researchers and engineers on ambiguous, fast-moving technical problems.\n\n - Clear written and verbal communication of technical ideas, results, tradeoffs, and risks.\n\n\nNICE TO HAVE\n\n - Experience with open-ended learning, automatic curriculum generation, intrinsic motivation, self-play, goal-conditioned RL, unsupervised skill discovery, multi-agent RL, quality-diversity, or evolutionary methods.\n\n - Familiarity with methods such as POET, population-based training, or multi-agent RL\n\n - Experience designing benchmarks or evals for generalization, exploration, long-horizon learning or behavioral diversity\n\n - Demonstrated taste for identifying non-obvious research directions and converting them into tractable experiments.\n\n\nROLE SPECIFIC LOCATION POLICY\n\n - this role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement\n\n\nCOMPENSATION\n\nThe expected salary range for this position is $300,000 - $500,000 USD"},{"id":"6d9e084c-b92c-427f-b30c-528cf91af02f","title":"Member of Technical Staff - Applied RL","department":"Research and Engineering","team":"Research and Engineering","employmentType":"FullTime","location":"SF","secondaryLocations":[],"publishedAt":"2025-11-05T20:29:03.729+00:00","isListed":true,"isRemote":null,"workplaceType":null,"address":null,"jobUrl":"https://jobs.ashbyhq.com/vmax/6d9e084c-b92c-427f-b30c-528cf91af02f","applyUrl":"https://jobs.ashbyhq.com/vmax/6d9e084c-b92c-427f-b30c-528cf91af02f/application","descriptionHtml":"Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.
This role is for exceptional ML engineers who can turn RL research ideas into working training systems, evals, environment and rewards. You will work across research and engineering to make post-training methods reliable, measurable, and fast to iterate on.
Build and improve RL training pipelines for language model based agents.
Translate research ideas into working implementations, including reward functions, verifiers, environment interfaces, rollout pipelines, and evaluation harnesses.
Design experiments that test whether RL methods are actually improving model behavior, sample efficiency, robustness, or generalization.
Create quality monitoring tools for RL experiments, including regression tests, eval suites, and reward-hacking checks.
Debug unstable training runs, diagnose poor learning dynamics, and identify whether failures come from algorithms, rewards, data, infrastructure, or evals.
Build 0→1 systems for new RL workflows, then harden them into reusable infrastructure.
Improve the reliability, reproducibility, and speed of experimentation across RL projects.
Own technically ambiguous projects end to end, from problem framing through implementation, evaluation, and iteration.
Strong practical ML engineering ability, demonstrated through shipped systems, open-source projects, competitions, independent projects, or equivalent experience.
Hands-on experience building, training, evaluating, or debugging ML systems.
Strong programming ability in Python and experience with at least one major ML framework such as PyTorch or JAX.
Working understanding of reinforcement learning, supervised learning, optimization, and modern deep learning.
Ability to independently take an ambiguous technical problem and drive it to a working implementation..
Ability to collaborate closely with researchers while maintaining high engineering standards.
Experience building systems that are reliable, maintainable, and usable by other technical team members.
Clear written and verbal communication.
Experience supporting research teams or fast-moving ML teams.
Expertise in building experiment tracking, evaluation platforms, dataset/versioning systems, or reproducibility infrastructure.
Experience at a high engineering bar organization where reliability, ownership, and code quality were central.
Experience reducing operational complexity in systems that had become brittle, slow, or hard to debug.
This role is based in our San Francisco office; For exceptional candidates we are willing to consider a hybrid arrangement
The expected salary range for this position is $300,000 - $500,000 USD
","descriptionPlain":"ABOUT VMAX\n\nVmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.\n\n\nABOUT THE ROLE\n\nThis role is for exceptional ML engineers who can turn RL research ideas into working training systems, evals, environment and rewards. You will work across research and engineering to make post-training methods reliable, measurable, and fast to iterate on.\n\n\nRESPONSIBILITIES\n\n - Build and improve RL training pipelines for language model based agents.\n\n - Translate research ideas into working implementations, including reward functions, verifiers, environment interfaces, rollout pipelines, and evaluation harnesses.\n\n - Design experiments that test whether RL methods are actually improving model behavior, sample efficiency, robustness, or generalization.\n\n - Create quality monitoring tools for RL experiments, including regression tests, eval suites, and reward-hacking checks.\n\n - Debug unstable training runs, diagnose poor learning dynamics, and identify whether failures come from algorithms, rewards, data, infrastructure, or evals.\n\n - Build 0→1 systems for new RL workflows, then harden them into reusable infrastructure.\n\n - Improve the reliability, reproducibility, and speed of experimentation across RL projects.\n\n - Own technically ambiguous projects end to end, from problem framing through implementation, evaluation, and iteration.\n\n\nMINIMUM REQUIREMENTS\n\n - Strong practical ML engineering ability, demonstrated through shipped systems, open-source projects, competitions, independent projects, or equivalent experience.\n\n - Hands-on experience building, training, evaluating, or debugging ML systems.\n\n - Strong programming ability in Python and experience with at least one major ML framework such as PyTorch or JAX.\n\n - Working understanding of reinforcement learning, supervised learning, optimization, and modern deep learning.\n\n - Ability to independently take an ambiguous technical problem and drive it to a working implementation..\n\n - Ability to collaborate closely with researchers while maintaining high engineering standards.\n\n - Experience building systems that are reliable, maintainable, and usable by other technical team members.\n\n - Clear written and verbal communication.\n\n\nNICE TO HAVE\n\n - Experience supporting research teams or fast-moving ML teams.\n\n - Expertise in building experiment tracking, evaluation platforms, dataset/versioning systems, or reproducibility infrastructure.\n\n - Experience at a high engineering bar organization where reliability, ownership, and code quality were central.\n\n - Experience reducing operational complexity in systems that had become brittle, slow, or hard to debug.\n\n\nROLE SPECIFIC LOCATION POLICY\n\n - This role is based in our San Francisco office; For exceptional candidates we are willing to consider a hybrid arrangement\n\n\nCOMPENSATION\n\nThe expected salary range for this position is $300,000 - $500,000 USD"},{"id":"c4e1e51b-65c5-41f0-9bed-c90bc452859b","title":"Research Fellowship - Open Endedness","department":"Research and Engineering","team":"Research and Engineering","employmentType":"FullTime","location":"SF","secondaryLocations":[],"publishedAt":"2025-11-05T20:31:18.089+00:00","isListed":true,"isRemote":null,"workplaceType":null,"address":null,"jobUrl":"https://jobs.ashbyhq.com/vmax/c4e1e51b-65c5-41f0-9bed-c90bc452859b","applyUrl":"https://jobs.ashbyhq.com/vmax/c4e1e51b-65c5-41f0-9bed-c90bc452859b/application","descriptionHtml":"Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.
A core focus of ours is agents that can learn to find their own objectives in the world. We are looking for researchers to design and build new ways of using RL where the formulation of rewards and tasks need to be discovered, rather than given.
This 3 to 6 month fellowship is for PhD students or equivalent early-career researchers who want to work on LLMs that can learn in open-ended settings. You will own a focused research project, work closely with Vmax technical staff, and contribute to research publications.
Develop RL methods for agents that can discover useful objectives, tasks and curricula without relying entirely on human-specified rewards.
Design systems for open-ended learning, including unsupervised/automated environment design, asymmetric self-play, and intrinsic motivation.
Build training loops where agents learn from interaction, exploration, novelty, competence progress, self-generated challenges, or other nonstandard reward signals.
Investigate how agents can avoid collapse into trivial, degenerate, or easily exploitable objectives.
Own and develop a research agenda within Vmax, from identifying promising directions to executing experiments and communicating results.
Currently enrolled in a PhD program in machine learning, computer science, artificial intelligence, computational neuroscience, mathematics, or a related technical field. Exceptional candidates with equivalent research experience may also be considered.
Track record of research excellence or strong research promise, demonstrated through publications, preprints, open-source work, technical projects, competitions, or publicly available artifacts.
Working understanding of reinforcement learning.
Familiarity with unsupervised/automated environment design, asymmetric self-play, and/or intrinsic motivation.
Strong programming ability in Python and experience with at least one major ML framework such as PyTorch or JAX.
Clear written and verbal communication of technical ideas.
Experience with LLM post-training methods
Experience with scalable ML experimentation, distributed training, experiment tracking, or reproducible research infrastructure.
Demonstrated taste for identifying non-obvious research directions and converting them into tractable experiments.
This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement
Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.
LLMs are fantastically powerful and there is a rapidly growing corpus of work devoted to understanding their internal representations and computations. We use the tools of mechanistic interpretability to enhance reinforcement learning by generating intrinsic rewards as a supplement or alternative to downstream human-generated verifiers.
Develop methods for using mechanistic interpretability to extract useful training signals from the internal states of language models.
Turn representations, features, circuits, and causal model behaviors into intrinsic rewards for reinforcement learning.
Compare interpretability-derived rewards against human feedback, learned reward models, verifiers, and task-level outcome rewards.
Design metrics and baselines for reward quality, including alignment with intended behavior, generalization across tasks, robustness, and resistance to reward hacking.
Investigate how internal representations evolve during RL and post-training, and use these insights to improve training objectives.
Develop infrastructure for reproducible, large-scale experiments on LLM agents, interpretability tools, and RL environments.
Define and pursue a high-impact research agenda that advances Vmax’s goal of open-ended learning beyond imitation of human expertise.
PhD or equivalent experience in machine learning, reinforcement learning, or a closely related field.
Track record of research excellence, as demonstrated by publications, open source work, deployed AI systems, or other substantial technical contributions.
Deep understanding of modern machine learning, especially reinforcement learning, representation learning, and large language models.
Strong familiarity with LLM post-training methods
Experience designing and running rigorous ML experiments, including ablations, baselines, evaluation design, and failure analysis.
Expertise with Python and at least one major ML framework such as PyTorch or JAX.
Ability to work independently on open-ended research problems and turn ambiguous ideas into concrete experimental programs.
Experience with mechanistic interpretability techniques such as activation patching, probing, sparse autoencoders, feature attribution
Experience training or evaluating language-model agents in interactive, tool-using, or multi-step reasoning settings.
Familiarity with scalable RL infrastructure, distributed training, experiment tracking, and large-scale evaluation pipelines.
Experience developing reward models, verifiers, process supervision methods, or automated evaluation systems.
Demonstrated software engineering ability, especially in research codebases that require reliability, reproducibility, and iteration speed.
Ability to present technical results and their strategic implications to both research and non-research audiences.
This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement
The expected salary range for this position is $300,000 - $500,000 USD
","descriptionPlain":"ABOUT VMAX\n\nVmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.\n\n\nABOUT THE ROLE\n\nLLMs are fantastically powerful and there is a rapidly growing corpus of work devoted to understanding their internal representations and computations. We use the tools of mechanistic interpretability to enhance reinforcement learning by generating intrinsic rewards as a supplement or alternative to downstream human-generated verifiers. \n\n\nRESPONSIBILITIES\n\n - Develop methods for using mechanistic interpretability to extract useful training signals from the internal states of language models.\n\n - Turn representations, features, circuits, and causal model behaviors into intrinsic rewards for reinforcement learning.\n\n - Compare interpretability-derived rewards against human feedback, learned reward models, verifiers, and task-level outcome rewards.\n\n - Design metrics and baselines for reward quality, including alignment with intended behavior, generalization across tasks, robustness, and resistance to reward hacking.\n\n - Investigate how internal representations evolve during RL and post-training, and use these insights to improve training objectives.\n\n - Develop infrastructure for reproducible, large-scale experiments on LLM agents, interpretability tools, and RL environments.\n\n - Define and pursue a high-impact research agenda that advances Vmax’s goal of open-ended learning beyond imitation of human expertise.\n\n\nMINIMUM REQUIREMENTS\n\n - PhD or equivalent experience in machine learning, reinforcement learning, or a closely related field.\n\n - Track record of research excellence, as demonstrated by publications, open source work, deployed AI systems, or other substantial technical contributions.\n\n - Deep understanding of modern machine learning, especially reinforcement learning, representation learning, and large language models.\n\n - Strong familiarity with LLM post-training methods\n\n - Experience designing and running rigorous ML experiments, including ablations, baselines, evaluation design, and failure analysis.\n\n - Expertise with Python and at least one major ML framework such as PyTorch or JAX.\n\n - Ability to work independently on open-ended research problems and turn ambiguous ideas into concrete experimental programs.\n\n\nNICE TO HAVE\n\n - Experience with mechanistic interpretability techniques such as activation patching, probing, sparse autoencoders, feature attribution\n\n - Experience training or evaluating language-model agents in interactive, tool-using, or multi-step reasoning settings.\n\n - Familiarity with scalable RL infrastructure, distributed training, experiment tracking, and large-scale evaluation pipelines.\n\n - Experience developing reward models, verifiers, process supervision methods, or automated evaluation systems.\n\n - Demonstrated software engineering ability, especially in research codebases that require reliability, reproducibility, and iteration speed.\n\n - Ability to present technical results and their strategic implications to both research and non-research audiences.\n\n\nROLE SPECIFIC LOCATION POLICY\n\n - This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement\n\n\nCOMPENSATION\n\nThe expected salary range for this position is $300,000 - $500,000 USD"},{"id":"87679e56-f98e-4e7d-a071-183444be8c48","title":"Member of Technical Staff - RL Algorithms","department":"Research and Engineering","team":"Research and Engineering","employmentType":"FullTime","location":"SF","secondaryLocations":[],"publishedAt":"2026-05-20T16:34:32.374+00:00","isListed":true,"isRemote":null,"workplaceType":null,"address":null,"jobUrl":"https://jobs.ashbyhq.com/vmax/87679e56-f98e-4e7d-a071-183444be8c48","applyUrl":"https://jobs.ashbyhq.com/vmax/87679e56-f98e-4e7d-a071-183444be8c48/application","descriptionHtml":"Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.
RL has become the de-facto method of post-training LLMs. We are limited by the sample efficiency of the current policy gradient algorithms in use today, and are looking for a talented researcher to weave together pre-LLM and post-LLM approaches to learning from experience.
Develop new RL algorithms for post-training language models.
Adapt ideas from pre-LLM reinforcement learning, such as model-based RL, temporal abstraction, and value-based learning, to modern LLM and agentic settings.
Establish empirical baselines and evaluation protocols for measuring sample efficiency, robustness, generalization, and reward exploitation in LLM RL.
Analyze failure modes of RL-trained models, including reward hacking, mode collapse, over-optimization, exploration failures, and distribution shift.
Collaborate with researchers working on environments, evals, interpretability, reward modeling, and infrastructure to turn algorithmic ideas into reliable training systems.
Own and develop a research agenda within Vmax, from identifying promising directions to executing experiments and communicating results.
PhD or equivalent experience in machine learning, reinforcement learning, or a closely related field.
Track record of research excellence, as demonstrated by publications, open source work, deployed AI systems, or other substantial technical contributions.
Deep understanding of modern machine learning, especially reinforcement learning, representation learning, and large language models.
Strong familiarity with LLM post-training methods.
Experience designing and running rigorous ML experiments, including ablations, baselines, evaluation design, and failure analysis.
Experience with large-scale ML infrastructure, distributed training, experiment tracking, data pipelines, and debugging unstable training runs.
Expertise with Python and at least one major ML framework such as PyTorch or JAX.
Ability to work independently on open-ended research problems and turn ambiguous ideas into concrete experimental programs.
Experience developing new RL algorithms or improving existing ones in domains such as robotics, games, simulated control, language models, or agents.
Experience with LLM pre-training.
Strong understanding of reward modeling, verifiers, process supervision, outcome supervision, or automated evaluation systems.
Demonstrated software engineering ability
Strong communication skills, especially the ability to explain algorithmic ideas, empirical results, and research implications to both technical and non-technical audiences
This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement
The expected salary range for this position is $300,000 - $500,000 USD
","descriptionPlain":"ABOUT VMAX\n\nVmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.\n\n\nABOUT THE ROLE\n\nRL has become the de-facto method of post-training LLMs. We are limited by the sample efficiency of the current policy gradient algorithms in use today, and are looking for a talented researcher to weave together pre-LLM and post-LLM approaches to learning from experience.\n\n\nRESPONSIBILITIES\n\n - Develop new RL algorithms for post-training language models.\n\n - Adapt ideas from pre-LLM reinforcement learning, such as model-based RL, temporal abstraction, and value-based learning, to modern LLM and agentic settings.\n\n - Establish empirical baselines and evaluation protocols for measuring sample efficiency, robustness, generalization, and reward exploitation in LLM RL.\n\n - Analyze failure modes of RL-trained models, including reward hacking, mode collapse, over-optimization, exploration failures, and distribution shift.\n\n - Collaborate with researchers working on environments, evals, interpretability, reward modeling, and infrastructure to turn algorithmic ideas into reliable training systems.\n\n - Own and develop a research agenda within Vmax, from identifying promising directions to executing experiments and communicating results.\n\n\nMINIMUM REQUIREMENTS\n\n - PhD or equivalent experience in machine learning, reinforcement learning, or a closely related field.\n\n - Track record of research excellence, as demonstrated by publications, open source work, deployed AI systems, or other substantial technical contributions.\n\n - Deep understanding of modern machine learning, especially reinforcement learning, representation learning, and large language models.\n\n - Strong familiarity with LLM post-training methods.\n\n - Experience designing and running rigorous ML experiments, including ablations, baselines, evaluation design, and failure analysis.\n\n - Experience with large-scale ML infrastructure, distributed training, experiment tracking, data pipelines, and debugging unstable training runs.\n\n - Expertise with Python and at least one major ML framework such as PyTorch or JAX.\n\n - Ability to work independently on open-ended research problems and turn ambiguous ideas into concrete experimental programs.\n\n\nNICE TO HAVE\n\n - Experience developing new RL algorithms or improving existing ones in domains such as robotics, games, simulated control, language models, or agents.\n\n - Experience with LLM pre-training.\n\n - Strong understanding of reward modeling, verifiers, process supervision, outcome supervision, or automated evaluation systems.\n\n - Demonstrated software engineering ability\n\n - Strong communication skills, especially the ability to explain algorithmic ideas, empirical results, and research implications to both technical and non-technical audiences\n\n\nROLE SPECIFIC LOCATION POLICY\n\n - This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement\n\n\nCOMPENSATION\n\nThe expected salary range for this position is $300,000 - $500,000 USD"},{"id":"025ba97e-9a7b-4644-82a5-cf83236e55c0","title":"Member of Technical Staff - RL Infrastructure","department":"Research and Engineering","team":"Research and Engineering","employmentType":"FullTime","location":"SF","secondaryLocations":[],"publishedAt":"2026-05-20T16:35:46.541+00:00","isListed":true,"isRemote":null,"workplaceType":null,"address":null,"jobUrl":"https://jobs.ashbyhq.com/vmax/025ba97e-9a7b-4644-82a5-cf83236e55c0","applyUrl":"https://jobs.ashbyhq.com/vmax/025ba97e-9a7b-4644-82a5-cf83236e55c0/application","descriptionHtml":"Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.
This role is for strong infrastructure engineers who can build the systems layer for RL at scale: distributed rollouts, training orchestration, inference, evals, data pipelines, observability, and reliability. You will create the durable platform that enables researchers and applied ML engineers to run, debug, and reproduce large-scale RL experiments.
Build infrastructure for distributed RL training and inference across thousands of GPUs
Improve the reliability, debuggability, and throughput of RL experiments.
Build interfaces that allow researchers and applied ML engineers to launch, inspect, compare, and reproduce experiments easily.
Own infrastructure projects end to end, from architecture and implementation through deployment, documentation, and long-term maintenance.
Identify and eliminate bottlenecks in training, rollout generation, eval execution, data movement, and cluster utilization.
Maintain engineering standards for RL infrastructure, including testing, observability, versioning, and reproducibility.
Strong software engineering experience.
Experience building infrastructure for LLM inference and/or RL training.
Experience with GPU clusters, distributed training, model serving, or high-throughput inference systems.
Familiarity with vLLM, SGLang and modern LLM-RL training frameworks
Strong understanding of system reliability, observability, testing, debugging, and performance optimization.
Ability to work closely with ML researchers and translate messy experimental workflows into durable infrastructure.
Experience building tools, platforms, or services used by other technical users.
Strong judgment around technical tradeoffs: when to prototype, when to harden, when to simplify, and when to redesign.
Clear written and verbal communication, especially around system design, operational risks, and engineering tradeoffs.
Experience supporting research teams or fast-moving ML teams.
Experience at a high engineering bar organization where reliability, ownership, and code quality were central.
Evidence of strong independent technical work, such as open-source projects, infrastructure projects, competitions, or substantial systems built from scratch.
Experience reducing operational complexity in systems that had become brittle, slow, or hard to debug.
This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement
The expected salary range for this position is $300,000 - $500,000 USD
","descriptionPlain":"ABOUT VMAX\n\nVmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimising beyond the local maxima of learning from human expertise.\n\n\nABOUT THE ROLE\n\nThis role is for strong infrastructure engineers who can build the systems layer for RL at scale: distributed rollouts, training orchestration, inference, evals, data pipelines, observability, and reliability. You will create the durable platform that enables researchers and applied ML engineers to run, debug, and reproduce large-scale RL experiments.\n\n\nRESPONSIBILITIES\n\n - Build infrastructure for distributed RL training and inference across thousands of GPUs\n\n - Improve the reliability, debuggability, and throughput of RL experiments.\n\n - Build interfaces that allow researchers and applied ML engineers to launch, inspect, compare, and reproduce experiments easily.\n\n - Own infrastructure projects end to end, from architecture and implementation through deployment, documentation, and long-term maintenance.\n\n - Identify and eliminate bottlenecks in training, rollout generation, eval execution, data movement, and cluster utilization.\n\n - Maintain engineering standards for RL infrastructure, including testing, observability, versioning, and reproducibility.\n\n\nMINIMUM REQUIREMENTS\n\n - Strong software engineering experience.\n\n - Experience building infrastructure for LLM inference and/or RL training. \n\n - Experience with GPU clusters, distributed training, model serving, or high-throughput inference systems.\n\n - Familiarity with vLLM, SGLang and modern LLM-RL training frameworks\n\n - Strong understanding of system reliability, observability, testing, debugging, and performance optimization.\n\n - Ability to work closely with ML researchers and translate messy experimental workflows into durable infrastructure.\n\n - Experience building tools, platforms, or services used by other technical users.\n\n - Strong judgment around technical tradeoffs: when to prototype, when to harden, when to simplify, and when to redesign.\n\n - Clear written and verbal communication, especially around system design, operational risks, and engineering tradeoffs.\n\n\nNICE TO HAVE\n\n - Experience supporting research teams or fast-moving ML teams.\n\n - Experience at a high engineering bar organization where reliability, ownership, and code quality were central.\n\n - Evidence of strong independent technical work, such as open-source projects, infrastructure projects, competitions, or substantial systems built from scratch.\n\n - Experience reducing operational complexity in systems that had become brittle, slow, or hard to debug.\n\n\nROLE SPECIFIC LOCATION POLICY\n\n - This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement\n\n\nCOMPENSATION\n\nThe expected salary range for this position is $300,000 - $500,000 USD"},{"id":"cafb6152-b8e5-4b5c-8a28-586607270996","title":"Open Application – Exceptional Talent","department":"Research and Engineering","team":"Research and Engineering","employmentType":"FullTime","location":"SF","secondaryLocations":[],"publishedAt":"2026-05-20T16:12:41.219+00:00","isListed":true,"isRemote":null,"workplaceType":null,"address":null,"jobUrl":"https://jobs.ashbyhq.com/vmax/cafb6152-b8e5-4b5c-8a28-586607270996","applyUrl":"https://jobs.ashbyhq.com/vmax/cafb6152-b8e5-4b5c-8a28-586607270996/application","descriptionHtml":"Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimizing beyond the local maxima of learning from human expertise.
Vmax is always interested in exceptional people who can help us build AI systems capable of open-ended learning. This application is for candidates whose strongest contribution may not fit neatly into one of our current roles, but who have clear evidence of unusually strong ability, agency, and or ambition.
Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimizing beyond the local maxima of learning from human expertise.
LLMs are fantastically powerful and there is a rapidly growing corpus of work devoted to understanding their internal representations and computations. We use the tools of mechanistic interpretability to enhance reinforcement learning by generating intrinsic rewards as a supplement or alternative to downstream human-generated verifiers.
This 3 to 6 month fellowship is for PhD students or equivalent early-career researchers who want to work at the intersection of mechanistic interpretability and reinforcement learning. You will own a focused research project, work closely with Vmax technical staff, and contribute to research publications.
Develop mechanistic interpretability methods for understanding internal representations, features, circuits, and computations in language models and agents.
Investigate how model internals can be used to generate intrinsic rewards, auxiliary objectives, diagnostics, or training signals for reinforcement learning.
Design and run experiments that test whether interpretability-derived signals improve learning, exploration, generalization, robustness, or sample efficiency.
Compare internally derived rewards against baselines such as human-generated verifiers, reward models, task-level outcome rewards, and standard intrinsic motivation methods.
Use techniques such as probing, activation analysis, sparse autoencoders, causal interventions, feature attribution, or representation analysis to study model behavior.
Analyze failure modes, including reward hacking, spurious features, non-causal correlations, objective misspecification, and overfitting to narrow evaluation distributions.
Build research code, evaluation harnesses, and experimental infrastructure that make results reproducible and useful to the broader team.
Communicate research progress clearly through written updates, internal presentations, and final project outputs.
Currently enrolled in a PhD program in machine learning, computer science, artificial intelligence, computational neuroscience, mathematics, or a related technical field. Exceptional candidates with equivalent research experience may also be considered.
Track record of research excellence or strong research promise, demonstrated through publications, preprints, open-source work, technical projects, competitions, or publicly available artifacts.
Working understanding of reinforcement learning.
Familiarity with mechanistic interpretability, representation analysis, or empirical methods for understanding neural networks.
Strong programming ability in Python and experience with at least one major ML framework such as PyTorch or JAX.
Clear written and verbal communication of technical ideas.
Experience with LLM post-training methods
Familiarity with intrinsic motivation, unsupervised RL, auxiliary objectives, representation learning for RL, or curiosity-driven learning.
Experience with scalable ML experimentation, distributed training, experiment tracking, or reproducible research infrastructure.
Interest in turning mechanistic understanding into practical training methods, rather than only analyzing models after training.
This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement