Research Scientist - Geo-Distributed Inference

Pluralis-Research; San Francisco (CA); Remote; USD 180,000 - 280,000; Full time

Benefits offered by this job

Equity-heavy package

Remote-first culture

Visa sponsorship & relocation

Open problems

Job summary

Pluralis Research is building Protocol Learning to train and serve large models on consumer hardware via the internet. The role focuses on developing a fast and reliable inference pipeline that can operate end-to-end across distributed nodes, while enabling production-grade serving for RL training and model deployment.

You will design and implement novel methods for fast inference on consumer GPUs, contribute to research publications or presentations, and operate in a remote-first environment

Qualifications

• Shipped serving engines or built a large-scale inference system.
• Publications (papers/blog posts) in distributed inference or LLM serving systems are a strong signal.
• Experience with systems in low-bandwidth, high-latency settings.
• Alignment with Protocol Learning mission is desirable.

Responsibilities

• Own the inference stack end-to-end: pipeline-parallel execution, placement, routing, transport, serving engine, and failure handling.
• Invent algorithms to accelerate inference on consumer hardware over the public internet and validate them in production.
• Serve training and users: keep rollout pipeline fast for RL training now and enable serving for trained models.

Skills

Serving systems

Research ability

Low-bandwidth networks

Mission alignment

Job description

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.

Our inference pipeline generates the rollouts for reinforcement learning (RL) training today, and it'll serve our models once they're trained. It also runs in a permissionless, trustless setting, which makes the usual serving problem much harder. The hardware is Macs and consumer GPUs owned by strangers, the network is the public internet, nodes join and leave mid-run, and the weights change under the server as training moves. Your primary role is to build the systems that keep this pipeline fast and reliable under these conditions.

Key Responsibilities

• Own the inference stack: You build and own it end-to-end. Pipeline-parallel execution, placement and routing, the transport, the serving engine, and failure handling. You set the direction, and you make things happen.

• Invent the algorithms: Making inference fast on consumer hardware over the public internet takes methods that don't exist yet. You design them, validate them, and put them in production.

• Serve training and users: You keep the rollout pipeline fast and reliable for RL training now, and turn it into the serving layer for our models once they're trained.

What We're Looking For

• Shipped serving systems: You've shipped serving-engine internals or built a large-scale inference system yourself, and you can do this work hands-on today.

• Research ability: Publications (papers and blogposts) in distributed inference or a nearby field, such as LLM serving systems, pipeline parallelism over slow networks, or decentralized training, are a strong signal. So is unpublished work you can walk us through.

• Low-bandwidth networking: Experience with systems that run in low-bandwidth, high-latency settings like the public internet is a strong signal.

• Mission alignment: You believe Protocol Learning is the viable third path for collective, trustless, and sovereign AI.

Nice to Have

• Familiarity with RL post-training.

• Exposure to Apple silicon or MLX.

• 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 this resonates, please apply.