Research Scientist - Geo-Distributed Inference
Pluralis Research · Remote · San Francisco · Australia ·
COMPENSATION
€•• – ••k
Listed on the employer's job board (Ashby)
- Fully remote worldwide
- Visa sponsorship available
- Senior research role
- Decentralized AI focus
- Professional English required
You will own the end-to-end inference stack for Protocol Learning, enabling training and serving of large models on distributed consumer devices. The work requires inventing new methods to handle low-bandwidth networks, dynamic node availability, and changing model weights in a trustless environment.
What you'll do
- Own the inference stack including pipeline execution and routing
- Design and validate new algorithms for low-bandwidth inference
- Ensure reliability of the rollout pipeline for RL training
- Build the serving layer for trained models on consumer hardware
- Handle failure scenarios in a permissionless network setting
What you'll need
- Experience shipping serving-engine internals or large-scale inference systems
- Publications or demonstrable work in distributed inference or LLM serving
- Expertise in systems running over low-bandwidth, high-latency networks
- Alignment with the mission of trustless and sovereign AI development
- Professional-level written and spoken English proficiency
Nice to have
- Familiarity with reinforcement learning post-training techniques
- Exposure to Apple silicon or the MLX framework
- Experience with P2P networking and NAT traversal
- Background at proprietary or open-source AI laboratories
Stack
- MLX
- Apple silicon
How you'll work
- Fully remote with team members distributed globally
- Main teams located in Australia and North America
- Comfort with working across multiple time zones required
What you get
- Equity-heavy package with significant ownership opportunities
- High base salary component
- Optional full visa sponsorship for Australia or US relocation
- Flexible work environment