On-Device AI Researcher — Lead Next-Gen Inference

Base Compute; Melbourne; Vor Ort; AUD 150.000 - 210.000; Vollzeit

Benefits dieser Stelle

Equity

Hybrid work (occasional in-office)

Direct influence on technical方向

Zusammenfassung

Base Compute, an on-device AI lab based in Melbourne, seeks an ML Researcher to push the frontier of on-device AI. You will identify problems, design experiments and derive insights that translate into real-world performance improvements.

You’ll own parts of the research agenda and influence the company’s technical bets. You’ll work on inference research, model routing between on-device and frontier models, and autoresearch pipelines to accelerate the R&D loop while building rigorous evaluations

Qualifikationen

• PhD in ML or equivalent industry research experience.
• Deep understanding of LLM architectures and AI inference.
• Expertise in speculative decoding, quantization theory, distillation, reinforcement learning.
• Proven track record: papers, open-source projects or blogs.
• Strong communication and reproducible documentation.

Aufgaben

• Identify and validate on-device efficiency approaches via experiments.
• Build and evaluate model routing between on-device and frontier API models.
• Design autonomous pipelines to explore and evaluate research ideas.
• Develop rigorous real-world performance evaluations and benchmarks.

Kenntnisse

PhD in ML

LLM architectures

Inference optimization

Research dissemination

Ausbildung

PhD in ML or equivalent

Tools

CUDA

ROCm

Metal

Triton

Jobbeschreibung

Base Compute, an on-device AI lab based in Melbourne, seeks an ML Researcher to push the frontier of on-device AI. You will identify problems, design experiments and derive insights that translate into real-world performance improvements.

You’ll own parts of the research agenda and influence the company’s technical bets. You’ll work on inference research, model routing between on-device and frontier models, and autoresearch pipelines to accelerate the R&D loop while building rigorous evaluations