Founding Member of Technical Staff - AI Research Scientist

h3bAbout GenPeach AI /b /h3pGenPeach AI is a product-driven research lab building vertical multimodal foundation models for hyper-realistic human generation in image and video – designed for emotionally resonant, human-centered AI experiences. Our goal is to create tools that supercharge human creativity rather than replace it. /ppbWe train models from scratch: /bproprietary datasets at massive scale, novel architectures and training recipes, large GPU clusters, and tight product integration so research ships to users quickly. /ppWe are a deeply technical team of around 10 people. We’re advised by Directors from Google DeepMind and backed by leading AI-focused funds and angels from OpenAI, Meta AI, Microsoft AI, Project Prometheus, and Fal. Collectively, our team, advisors, and angels have contributed to models including Meta’s Imagine/MovieGen and foundation-model work behind OpenAI’s Sora, plus Google’s Veo and Gemini. /ph3bAbout the Role /b /h3pWe’re looking for an exceptional bAI Research Scientist /b to help lead the next generation of GenPeach AI foundation models. This is a bLeadership IC /b role: you’ll shape research direction through hands-on experimentation, guide technical decisions across the stack, and mentor a small group of researchers while working closely with the founders. /ppThis isn’t “fine-tuning an open-source checkpoint.” It’s building new capabilities through architecture work, training systems, and post-training recipes—then deploying them into production. /ph3bIn this role, you will /b /h3ullipCo-design and train large-scale diffusion models for image and video generation /p /lilipBuild and iterate on btraining recipes /b (pretraining, post-training, control, preference/tuning where relevant) to unlock new model capabilities /p /lilipRun rigorous ablations: isolate what works, why it works, and communicate outcomes clearly to drive roadmap decisions /p /lilipReason about speed/quality/cost tradeoffs and make technical choices that materially affect training efficiency and production quality /p /lilipInfluence and contribute to dataset strategy: curation signals, filtering, evaluations, and feedback loops from product /p /lilipCollaborate with engineering/product to productionize research (serving constraints, stability, monitoring, fast iteration) /p /lilipMentor and raise the bar for a small team of researchers through code, reviews, and research hygiene /p /li /ulh3bYou might thrive in this role if you /b /h3ullipHave b5+ years /b of deep learning / applied AI research experience (or equivalent research impact) /p /lilipAre strong in bPython /b and bPyTorch /b, and comfortable owning research code that becomes production-critical /p /lilipHave hands-on experience training and debugging foundation models (not “black-box use”): you’ve dealt with instability, collapse, data issues, scaling pathologies, and know how to fix them /p /lilipCan move fast with good taste: you prioritize the experiments that matter and make decisions with incomplete information /p /lilipCommunicate research clearly – through writing, plots, ablations, and crisp takeaways /p /lilipTake ownership beyond your job description when needed (startup reality) /p /li /ulh3bMinimum Qualifications /b /h3ullip5+ years of experience in AI research / deep learning (industry or academia) /p /lilipExcellent grasp of modern generative modeling (e.g., diffusion/flow, GANs, VLMs or adjacent modalities) /p /lilipStrong software skills: Python + PyTorch, solid engineering hygiene for experimentation and reproducibility /p /lilipProven ability to drive projects end-to-end with high autonomy /p /li /ulh3bPreferred Qualifications /b /h3ullipExperience training blarge diffusion/flow models /b for image/video, or adjacent large-scale generative work (LLM/VLM/speech) with transferable scaling and post-training expertise /p /lilipExperience training at scale (multi-node, hundreds of GPUs), and understanding distributed training failure modes /p /lilipExperience designing model architectures or core training components (losses, conditioners, schedulers, sampling, distillation, etc.) /p /lilipPublications at top venues (NeurIPS/ICML/ICLR/CVPR/ICCV/ECCV/ACL) or equivalent research impact /p /lilipExperience leading technically (as an IC): mentoring, setting research direction, improving team execution /p /li /ulh3bWhat makes this role unique /b /h3ullipBuild frontier image/video models end-to-end: data to architecture to training to post-training to production /p /lilipHigh ownership and fast iteration in a lean team—your work directly shapes what we ship /p /lilipCollaboration across research and engineering with minimal process overhead /p /lilipA chance to compete on the global stage of foundation model quality—and ship results publicly /p /li /ulh3bHow we work /b /h3ullipHigh ownership and accountability /p /lilipDirect, low-ego communication /p /lilipBias toward impact: measure to iterate to ship /p /lilipStrong technical standards and research discipline /p /li /ulh3bLogistics /b /h3ullipLocation: bZurich (Switzerland) or Warsaw (Poland) /b — onsite or hybrid. If you’re elsewhere, we’re open to remote (team/timezone fit considered). /p /lilipCompensation: competitive salary + meaningful equity (level-dependent) /p /lilipInterview process: quick screen to research round to technical round (practical + systems) to team fit/values /p /li /ulh3bWhat we offer /b /h3ullipVisa sponsorship (where applicable); we’ll make a strong effort to relocate you to Switzerland or Poland if desired /p /lilipRemote-friendly: work fully remote, hybrid, or on-site from our hubs /p /lilipRegular offsites and in-person events to collaborate and connect /p /lilipFlexible PTO /p /li /ul #J-18808-Ljbffr