Senior AI Engineer (Trust & Safety)

Dreamerlabs; Sydney; On-site; AUD 180,000 - 240,000; Full time

Benefits offered by this job

Above-market pay

Performance raises

Growth opportunities

Meals provided

Company card & budget

North Sydney office

Team socials

Visa sponsorship

Job summary

Dreamerlabs in Sydney is seeking a senior engineering leader to own AI models across moderation, chat, image and voice capabilities. You will define what to build, train and deploy on GPU infrastructure, and measure accuracy, latency and cost at scale.

You will lead end‑to‑end ownership, partner with founders, and steer product direction while maintaining high standards for safety and performance in a fast‑growing environment.

Qualifications

• You’ve built something impressive and shipped it at real scale.
• Product judgement and stakeholder alignment matter.
• End-to-end ownership from idea to post‑launch.
• Trust and safety experience is a plus.
• Rigour, pragmatism and data‑driven decision making.
• Clear communication and collaborative mindset.
• Curiosity about AI and fast learning.

Responsibilities

• Own the AI models that power Dreamerlabs products and safety.
• Define detection requirements, train and evaluate on GPUs, deploy to production.
• Measure impact, latency and cost, iterate improvements.
• Balance accuracy with scalable, responsible moderation.
• Collaborate with founders and a world‑class engineering team.

Skills

Staff/Principal/Distinguished Engineer

End-to-end ownership

Strong communication

Curiosity about AI

Tools

Python

TypeScript

Job description

Dreamerlabs is building a new category of entertainment: interactive experiences powered by AI characters, chat, voice, image and video, shaped in real time around each person.

Entertainment has spent decades becoming more personalised, from TV to YouTube to TikTok. AI unlocks a second axis, interactivity, where the story responds to you. That is what we're building, and we're doing it at serious scale.

In under two years, we have built one of the most-visited AI entertainment products in the world:

• 20 million monthly visitors
• A top 1,000 website globally
• 250 billion tokens processed every day
• 1.2 million images and videos generated every day
• 350+ model requests per second, served from our own GPU fleet

We've done it profitable and bootstrapped, and now we are scaling our engineering team as we launch new products on the platform.

Why engineers join us

We fine-tune and deploy models for chat, image, video and voice, build our own models in-house, and run hundreds of GPUs (H100s and B200s) across multiple data centres. We are one of very few companies in Australia that work this close to the metal on generative AI.

We're AI-first in how we build, too. Everyone works with the latest AI coding tools, and we're constantly experimenting with how great engineers can do more than ever with them.

The role

You’ll own the models that keep our products safe, end to end. You’ll define what we need to catch, work with our labellers on the data, train and evaluate the models on our GPUs, and deploy them into production. You will then measure whether they worked and keep improving them.

You decide what to build and how, from traditional classifiers to LLMs, and you’re accountable for accuracy, latency and cost on models running across millions of images and messages a day.

Problems you could own:

• Image moderation. Improve detection of prohibited content across user prompts and generated images, cutting false positives without letting more through.
• Chat moderation. Use LLMs and classifiers to catch prohibited content in user messages and AI responses.
• The retraining loop. Get retrain, evaluate, shadow and deploy down to a week or less for any moderation model.
• Automating review. Automate parts of content review and user banning, with the right safeguards in place.

We're deliberately flat, with no levels and no management layers. The team includes former Heads of Engineering, Principal Engineers and founders who chose to be individual contributors because they love building, alongside engineers from Meta, Canva, Alibaba, Immutable, Relevance AI and Rokt. You'll work directly with the founders.

What we’re looking for

We care more about what you’ve built and how you think than years of experience or a checklist of technologies. You might be a Staff, Principal or Distinguished Engineer, a tech lead who has shaped product direction, a former Head of Engineering, or a founder who wants to get back to building.

• You’ve built something impressive. A technically complex product at real scale, or a hard problem you went deep on. You can explain clearly what you did and why it was hard.
• Product judgement. You talk to users, form a view on what's worth building, and back it with evidence. Ideally you’ve owned a customer-facing product, not only internal platforms.
• End-to-end ownership. You turn an ambiguous problem into a shipped product without someone breaking it into tasks, and you own what happens after launch.
• Trust and safety experience. Ideally you’ve built moderation or safety models at scale, and you know how to handle sensitive data properly.
• Rigour and pragmatism. You ask the next question (which users, how much impact, at what cost), run the experiment and check the result. You know when to build properly and when the simple solution is right.
• Clear communication. You explain your thinking succinctly, with data, and challenge ideas constructively.
• Curiosity about AI. You keep up with new models and tools and quickly see how they could change the product.
• Hunger. You want to win, hold yourself to high standards, and want the responsibility of owning something important.

Our product stack is TypeScript, Next.js, React, tRPC, Postgres, ClickHouse, Temporal and AWS. Our inference stack uses Python to serve LLMs, TTS, and diffusion image and video models on self-hosted GPUs. You don't need experience with all of it.

You’ll stand out if you've done some of this:

• Image moderation. Vision classifiers or VLMs tuned for precision at very high recall on prohibited content
• Text moderation. Distilling LLM labellers into small fine-tuned models for policy detection
• Labelling. LLM-assisted labelling and active learning so human reviewers only see the hard cases
• Retraining loop. Retraining classifiers on new abuse patterns within days using synthetic data and evals on real traffic
• Adversarial robustness. Catching users who gradually steer a conversation toward prohibited content

How we work

We work together in person in North Sydney, because things change fast and the best decisions happen in the room. We prioritise ruthlessly, test ideas in a day rather than planning them for weeks, and keep meetings and approvals to a minimum.

We think of ourselves as a competitive sports team with a real shot at winning. That means a high‑intensity environment, and it's built for people who want exactly that.

What you’ll get

• Above‑market pay. We reward exceptional people with above‑market salaries and meaningful equity.
• Fast reward for impact. Compensation is reviewed every six months, with raises and bonuses tied to what you deliver, not your title.
• Growth without management. Your scope, influence and pay can keep growing while you stay hands‑on.
• All meals covered. Breakfast, lunch and dinner every day, plus stocked fridges.
• A company card and unlimited workspace budget. For food, coffee, software, AI tools, equipment and the setup you need to do your best work.
• A great office. Dog‑friendly, in North Sydney, with a big balcony for working outdoors.
• Optional team socials. Gym sessions, running, swimming, escape rooms and more.
• Visa sponsorship. For engineers relocating to Australia.