Research Engineer - Post-Training

We need to output HTML fragment with sections: About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule as appropriate. Must only include sections that have real specific content. We must not repeat job title. Must not include Job ID etc. Must remove boilerplate. We need to parse the posting. Key parts: - Intro about Pluralis Research, Protocol Learning, Agora, post-training, role description. - Key Responsibilities: bullet list. - What We're Looking For: bullet list (hands-on RL post-training, strong engineering, research ability, mission alignment). Also Nice to Have list. - Compensation & Benefits: bullet list (Equity-Heavy Package, Remote-First Culture, Visa Sponsorship, Open Problems). Also FYI's: remote work, timezone, English proficiency, recruiters note, backing, ideological note. We need to map to sections. About the role: could be a paragraph summarizing role. Responsibilities: list from Key Responsibilities. Requirements: maybe combine What We're Looking For (mandatory) and Nice to Have? Usually Requirements = mandatory qualifications. Qualifications could be similar. But we need to decide. We have sections: About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule. We need to include only those with real content. We have content for About the role (intro). Responsibilities (list). Requirements (mandatory: hands-on RL post-training, strong engineering, research ability, mission alignment). Qualifications maybe same as Requirements? Could treat Qualifications as education/experience? Not given. Could skip Qualifications if no specific content. Skills: maybe list technical skills: Python, PyTorch, concurrency, failure handling, profiling, RL algorithms, etc. Could extract from responsibilities and requirements. Benefits: Equity-Heavy Package, Remote-First Culture, Visa Sponsorship, Open Problems (maybe not a benefit but could be). Also maybe "FYI's" includes remote work, timezone, English proficiency, etc. But those are more about work environment. Pay: Not explicitly stated salary numbers, only "high base salary" and equity. No specific numbers. The rule: never create Pay/Schedule sections unless posting actually states them. It says "high base salary" but no number. Is that considered stating? It's vague but still states there is a high base salary. Might be considered stating compensation. However the instruction: "never create Pay/Schedule/Benefits sections unless the posting actually states them." It says "high base salary" is a statement about compensation, albeit not numeric. I'd include Pay section with that statement. Schedule: Not mentioned; only remote-first culture, flexible work environment, timezone. No explicit schedule like full-time, part-time. So we should not create Schedule section. Benefits: we have Equity-Heavy Package, Remote-First Culture, Visa Sponsorship, Open Problems (maybe not benefit). Also "FYI's" includes remote work, timezone, English proficiency, etc. Could treat as Benefits. We need to avoid filler like "commensurate with experience". Not present. We must preserve all substantive information; do not invent or summarize away detail. We need to output HTML fragment only, no markdown fences. Let's craft. Structure:

About the role

Pluralis Research is developing Protocol Learning, a decentralized approach to train and serve large models on consumer-grade devices over the internet. Following the Agora run that pretrained an 8B model without any participant holding the full weights, the role focuses on making reinforcement‑learning post‑training work in this setting: building the end‑to‑end RL training loop, inventing algorithms tolerant of asynchronous, high‑latency, partially trusted rollouts, and shipping the first decentralized post‑trained models.

Responsibilities

• Build the post‑training stack: end‑to‑end RL training loop – rollout ingestion from the geo‑distributed inference pipeline, reward computation, policy updates, and distributing updated weights back to the network.
• Set direction and drive execution.
• Invent algorithms: adapt standard RL recipes to asynchronous, high‑latency, partially trusted generation (staleness tolerance, off‑policy corrections, communication‑efficient policy updates).
• Ship first post‑trained models: build evaluations that demonstrate improvement and take the first decentralized post‑trained release from run to public artifact.

Requirements

• Hands‑on experience with RL post‑training on large language models (RLHF, RLVR, or reasoning‑focused RL) and familiarity with the systems layer (rollout generation, async training loops, weight synchronization).
• Strong engineering: production‑quality Python and PyTorch, including concurrency, failure handling, and profiling before optimizing.
• Research ability: publications in RL post‑training, asynchronous or distributed RL, or related fields (or unpublished work you can defend in detail).
• Mission alignment: belief that Protocol Learning is a viable third path for collective, trustless, and sovereign AI.

Skills

• Python and PyTorch development.
• Concurrency and failure handling.
• Profiling and performance optimization.
• RL algorithms (RLHF, RLVR, reasoning‑focused RL).
• Asynchronous and distributed training techniques.
• Experience with rollout generation and weight synchronization.
• Familiarity with serving‑engine internals such as vLLM or SGLang (nice to have).
• Reward modeling or building verifiable‑reward datasets (nice to have).
• P2P networking and NAT traversal (nice to have).
• Experience training over slow networks or in decentralized/federated setups (nice to have).

Benefits

• Equity‑heavy package: significant ownership for key technical contributors plus a high base salary.
• Remote‑first culture: flexible work environment with team members distributed globally.
• Visa sponsorship: optional full visa sponsorship and relocation support to Australia or the US.
• Open problems: opportunity to write some of the first solutions for training and serving frontier models on uncontrolled hardware and networks.

Pay

High base salary combined with an equity‑heavy package.

We must not include Schedule. Check for any repeated content: we have covered all substantive info. We removed boilerplate: navigation, cookie, EEO, etc. We didn't include recruiters note, etc. We omitted "FYI's" about timezone, English proficiency, etc. Those are substantive? They mention timezone comfort and English proficiency. Could be considered requirements. But we already have mission alignment etc. Might be worth adding to Requirements or Skills. However the instruction: preserve ALL substantive information; do not invent or summarize away detail. So we should include timezone and English proficiency. Where to put? Could be under Requirements as additional items: "Must be comfortable working across timezones" and "Professional-level English proficiency (written and spoken)". Also "Applicants must have professional-level English proficiency". Also "Recruiters: we aren't looking for agency support at this time." That's boilerplate? It's a note to recruiters, not relevant to applicant; could be omitted as it's not substantive for the role. The "We'll reach out if we need help." also not needed. Also "We are backed by Union Square Ventures..." is background, could be included in About the role maybe. But we already have about role. Could add a sentence about backing. Also "Pluralis is unapologetically ideological..." could be part of mission alignment. We need to ensure we don't miss any substantive details. Let's list all substantive info from posting: - Pluralis Research works on Protocol Learning: training and serving large models in a fully decentralized way on small consumer-grade devices connected via the internet. - Despite being dismissed as infeasible, we have made significant advances on this problem, most recently Agora, a permissionless run that pretrained an 8B model from scratch on consumer GPUs spread over the internet, with no single participant ever holding the full weights (tech report). - While many of the core research problems have been solved, Protocol Learning unlocks a series of new challenges. - For the mission in full, read A Third Path: Protocol Learning. - Agora gave us a pretrained 8B model. Post-training is how we make it useful for agentic use-cases. - But every post-training stack you've seen assumes a datacenter — synchronous rollouts, fast interconnects, trusted workers. Ours gets none of that. It has to run on consumer GPUs, and Macs spread across the public internet, training a model whose weights no single participant ever holds, with rollouts arriving from a geo-distributed inference pipeline at high latencies. - Your primary role is to make RL post-training work here anyway — the algorithms and the system, end-to-end. Key Responsibilities: - Build the post-training stack: You build the RL training loop end-to-end: rollout ingestion from the geo-distributed inference pipeline, reward computation, policy updates, and getting updated weights back out to the network. - You set the direction, and you make things happen. - Invent the algorithms: Standard RL recipes assume on-policy rollouts from fast, trusted hardware. You adapt them to asynchronous, high-latency, partially trusted generation: staleness tolerance, off-policy corrections, and communication-efficient policy updates. - Ship first post-trained models: You build the evals that show the models are improving, and you take the first decentralized post-trained release from run to public artifact. What We're Looking For: - Hands-on RL post-training: You've run RL post-training on large language models — RLHF, RLVR, or reasoning-focused RL — and touched the systems layer yourself: rollout generation, async training loops, weight synchronization. Not just launched jobs on someone else's stack. - Strong engineering: Production-quality Python and PyTorch: concurrency, failure handling, profiling before optimizing. - Research ability: Publications in RL post-training, asynchronous or distributed RL, or nearby fields are a strong signal. So is unpublished work you can defend in detail. - Mission alignment: You believe Protocol Learning is the viable third path for collective, trustless, and sovereign AI. Nice to Have: - Experience training over slow networks, or with decentralized or federated setups. - Familiarity with serving-engine internals such as vLLM or SGLang — our rollout pipeline is a serving system. - Experience with reward modeling or building verifiable-reward datasets. - Experience with P2P networking and NAT traversal. - Experience at proprietary, open-weight and open-source AI labs Compensation & Benefits: - Equity-Heavy Package: We offer significant ownership for key technical contributors in addition to a high base salary. - Remote-First Culture: Flexible work environment with team members distributed globally. - Visa Sponsorship: Optional full visa sponsorship and relocation support to either Australia or the US. - Open Problems: Training and serving frontier models on hardware you don't control, over networks you don't own, mostly has no published answers yet. You'll write some of the first ones. FYI's: - We work remotely across the world, with the main teams in Australia and North America. You'll need to be comfortable working across timezones. - Applicants must have professional-level English proficiency (written and spoken). - Recruiters: we aren't looking for agency support at this time. We'll reach out if we need help. - We are backed by Union Square Ventures and other tier-1 investors, and we are a world-class, deeply technical team of ML researchers. - Pluralis is unapologetically ideological. We believe AI, and the world, end up on a better path if we succeed in implementing the protocol for intelligence. If