SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Pluralis Research is building Protocol Learning: a system for training and serving large language models in a fully decentralized manner on consumer-grade devices connected via the internet. The company has demonstrated feasibility with Agora, a permissionless run that pretrained an 8B model from scratch on consumer GPUs distributed across the internet, with no single participant holding full weights.
As Research Engineer for Geo-Distributed Inference, you will own the end-to-end inference stack that powers both the reinforcement learning training pipeline today and will serve models in production. This is a hands-on role where you design and build systems that operate in a permissionless, trustless setting with unique constraints: hardware consists of Macs and consumer GPUs owned by strangers, the network is the public internet, nodes join and leave mid-run, and model weights change during training.
Key responsibilities include: architecting and owning the complete inference pipeline (pipeline-parallel execution, placement and routing, transport layer, serving engine, and failure handling); inventing novel algorithms to make inference fast on consumer hardware over public internet; and maintaining reliability for both training rollouts and eventual user-facing model serving.
You should have shipped serving-engine internals or built large-scale inference systems, with hands-on capability to execute immediately. Research publications or unpublished work in distributed inference, LLM serving systems, pipeline parallelism over slow networks, or decentralized training are strong signals. Experience with low-bandwidth, high-latency systems like the public internet is valuable. Nice-to-have skills include RL post-training familiarity, Apple silicon/MLX exposure, P2P networking and NAT traversal experience, and background at proprietary, open-weight, or open-source AI labs.
The company operates remotely across the world with main teams in Australia and North America. You must be comfortable working across timezones and have professional-level English proficiency. Pluralis is backed by Union Square Ventures and tier-1 investors, with a deeply technical team of ML researchers focused on implementing a protocol for intelligence.
