Senior ML Research Engineer (Generative Media)

About AI Studio & The Lab

AI Studio puts every great AI model in one app, on iPhone, iPad and the web: cinematic video, 4K images, songs and voiceovers, actively used by over 10,000 creators every month and by more than 1,000,000 creators to date. We are self-funded with no outside investors, we ship every week, and decisions get made in days, not quarters. The company is registered in Türkiye and the team works remotely from Germany, Ukraine, Canada, Italy and Türkiye. The Lab is our in-house research group: a small team of ML engineers and data scientists building the proprietary models behind the app. Research here is measured by one thing: whether it makes what creators export better.

The Role

You will own our custom model post-training program. Our flagship tools; photo restoration, virtual try-on, consistent characters, signature styles, currently run on general-purpose frontier models. Your job is to replace them, one by one, with our own: fine-tunes and adapters of open-weight models (the Flux / Wan / Qwen class), trained on datasets we curate in-house, evaluated in our own harness, and served in production. You will work directly with the founders and define how post-training is done here.

• Type: Full-time
• Location: Fully remote; the team is spread across Germany, Ukraine, Canada, Italy and Türkiye
• Contract: Freelance / contract, long-term collaboration intended
• Compute: Dedicated GPU budget; you spec what the work needs

What You Will Do

• Post-train open-weight models: LoRAs, adapters, and full fine-tunes of state-of-the-art open image and video models; for restoration, try-on, identity-consistent characters, and style specialization.
• Build the datasets: Design curation, captioning, filtering, and augmentation pipelines for the proprietary training data behind each specialized model.
• Distill for speed: Work with the efficiency program on few-step distilled variants that power near-instant previews.
• Evaluate rigorously: Measure every checkpoint against the frontier in our internal harness; win rates, category breakdowns, regressions, before anything ships.
• Ship to production: Take your best checkpoints through serving and into the app, then watch real usage data tell you what to train next.
• Track the field: Evaluate every meaningful open-weight release within days and decide what it changes for our roadmap.

What We Are Looking For

Must-Haves

• 3+ years of hands-on deep learning experience, with real training runs you owned; not only inference or prompting.
• Direct experience training or fine-tuning diffusion / flow-matching models: LoRA and adapter methods, SFT, and preference optimization (DPO or similar).
• Strong PyTorch and the engineering skill to run multi-GPU training jobs unattended; mixed precision, checkpointing, data loading at scale.
• Experience building training datasets: curation, captioning, deduplication, and quality filtering.
• Evidence you can take a model from experiment to production, not just to a demo notebook.
• Good written English for async collaboration and clear experiment write-ups.

Nice-to-Haves

• Video diffusion experience (temporal layers, image-to-video conditioning, frame interpolation).
• Distillation experience; consistency models, adversarial or distribution-matching distillation.
• Familiarity with the open-weight ecosystem and its tooling (diffusers, ComfyUI internals, training frameworks).
• Publications or strong open technical write-ups in generative modeling.
• CUDA / performance engineering depth.

How We Work

• Research with a scoreboard: Every model you train is measured against the frontier and by real creators, no vanity benchmarks.
• Weeks, not quarters: The distance from a promising checkpoint to production is days. You will see your work used at scale immediately.
• Small team, full ownership: You are the post-training program. Direct line to the founders, no research bureaucracy.
• Competitive compensation: We value senior research talent and structure offers accordingly.

If You Are Interested

Use the form below. Alongside your CV, we would love to see:

• Your CV and LinkedIn profile, plus a short introduction.
• The training work you are proudest of; what you trained, on what data, and how you measured it. Links to write-ups, repos, or papers welcome.
• One open-weight image or video model you would post-train for our app first, and why.
• Your expected compensation for this position.

We read every application and reply to all of them.