Systems & Research Engineer | Applied AI | Up to A$420K + Equity | 100% Remote from Australia | US AI Company | Fully Async

Location: Fully remote, anywhere in the world. There is a San Francisco office if you want it, but attendance is not required.
Type: Full-time
Salary: USD $220k to $300k total compensation + equity

“I’ve seen things you people wouldn’t believe.” Blade Runner

Now we would like to see what you have built.

The company

We are partnering with a fast-growing applied AI company building serious production systems for the global freight and supply chain industry.

Founded by engineers from MIT and Stanford, the company raised USD $4.5M in seed funding in January 2026 and is already operating AI products at meaningful scale. One of its core fraud and identity platforms screens approximately 5,000 drivers every day.

This is not a giant engineering organisation adding another seat. It is a small, exceptionally ambitious technical team looking for engineers who can materially raise the bar.

The role

This is a Systems and Research Engineer position focused on the infrastructure, performance and research behind production AI systems.

Your world will include GPU and CPU performance, inference, model serving, evaluation, distributed systems, concurrency, latency, cost, benchmarking and performance engineering.

You will investigate how AI systems behave under real production conditions, identify bottlenecks, design experiments and make engineering decisions based on evidence rather than instinct.

A recent example involved benchmarking speech to text approaches and developing a hybrid open source and production architecture that beat alternative vendors on both quality and economics. That is the kind of problem you will be solving.

We care about the experiment

This may be the most important section of the advert.

We do not just want to hear:

• “Reduced latency by 35%.”
• “Built scalable ML infrastructure.”
• “Improved GPU utilisation.”

Those numbers are interesting. But what was the baseline? What hypothesis were you testing? Why did you design the benchmark that way? What alternatives did you compare? What did the data actually show?

And most importantly, what decision changed because of the experiment?

You need to be able to walk us through at least one rigorous benchmark or experiment you personally designed and ran involving model serving, inference, evaluation, retrieval, speech systems or agent infrastructure.

This team values engineers who think like researchers and researchers who can ship production systems.

What you might work on

• Profiling production AI systems and finding GPU, CPU, memory or network bottlenecks
• Benchmarking frameworks such as vLLM, SGLang or alternative serving approaches
• Testing inference throughput and latency
• Evaluating open source versus closed source models
• Optimising AI workloads for cost at high concurrency
• Building evaluation infrastructure
• Investigating model behaviour under production traffic
• Reasoning about distributed systems where a model is actually in the loop
• Turning research findings into production architecture decisions
• Using coding agents extensively to move faster while keeping engineering judgement firmly human

Who we are looking for

You might currently be a:

• Research Engineer
• ML Systems Engineer
• AI Infrastructure Engineer
• Inference Engineer
• Performance Engineer
• ML Platform Engineer
• Systems Engineer

Big tech ML infrastructure experience is highly relevant. Think environments like Google, Meta, Stripe, Amazon AGI or similarly sophisticated AI organisations.

But pedigree alone will not get you through. We are looking for evidence of exceptional technical work. The strongest candidates can explain their work like this: here was the hypothesis, here was the baseline, here was the experiment, here is what surprised us, here is the decision we made.

That is far more valuable than a CV packed with technical terminology and percentages without context.

What this is not

This is not a traditional DevOps or cloud infrastructure role. It is not frontend. It is not full stack. It is not Web3.

And sophisticated distributed systems experience alone is not enough if there is no model in the loop. The research and performance work needs to involve real AI systems.

AI coding agents

This team uses coding agents heavily every day. The expectation is that excellent engineers increasingly spend their time deciding what to build, how to test it and whether the result is correct, rather than manually writing every line. Fluency with modern coding agents will be assessed during the interview process.

Truly remote

This role is fully remote, anywhere in the world. San Francisco. New York. London. Singapore. Berlin. Toronto. Sydney. Your geography matters far less than the quality of your engineering.

The company works asynchronously across global time zones, so strong written communication and genuine ownership matter.

Compensation

USD $220k to $300k total compensation plus equity. USD $300k is the ceiling, and we are looking for engineers already operating comfortably within that range.

The bar

The founders have one overriding priority: hire engineers who make the existing engineering team better. That means the bar is deliberately high.

If you have done genuinely exceptional work around AI systems, inference, model serving, evaluation or performance engineering, we would like to hear about it.

When you apply, do not just tell us what you built. Tell us about the experiment. What did you believe? What did you measure? What did you discover? And what did you change because of it?