机器学习工程师（扩散模型 / 视觉方向） Machine Learning Engineer (Diffusion/Vision) at Bjak

Yuexiu District, Guangdong Province, China -

Full Time

Start Date

Immediate

Expiry Date

27 Dec, 25

Salary

0.0

Posted On

28 Sep, 25

Experience

2 year(s) or above

Remote Job

Yes

Telecommute

Yes

Sponsor Visa

Skills

Machine Learning, Diffusion Models, Visual Generation, Data Annotation, Dataset Preparation, PyTorch, GPU Optimization, Performance Optimization, Software Engineering, LoRA, DreamBooth, Stable Diffusion, SDXL, Flux, NSFW Detection, Adversarial Testing

Industry

Software Development

Description

机器学习工程师（扩散模型 / 视觉方向）将视觉模型转化为现实应用我们正在为全球用户构建 AI 系统。当前正处于 AI 变革时代 —— 本项目团队专注于打造真正具有现实世界影响力、全球高使用率的视觉生成类 AI 应用。该职位是一个全球岗位，支持灵活混合办公模式 — 结合远程办公与总部面对面协作。你将与产品、工程、运营、基础设施与数据等地区团队紧密协作，构建并扩展具有深远影响的 AI 解决方案。为什么这个岗位重要你将对最前沿的扩散模型进行微调，构建评估体系，并推动 AI 功能上线至生产环境。你的工作不仅让模型更智能，也确保其具备安全性、可信度与大规模应用能力。你的职责微调与定制 —— 使用 LoRA、DreamBooth 等参数高效方法训练并优化扩散模型（包括 SDXL、Flux、Stable Diffusion 等变体）数据集构建 —— 构建、清洗和标注大规模图像数据集，涵盖自动生成描述、图像打标签、NSFW 过滤等，以确保生成内容的安全性与对齐性评估与对齐 —— 设计评估流程，衡量生成图像在保真度、多样性、风格一致性与安全性等维度的表现性能优化 —— 应用 GPU 显存优化、潜空间扩散技巧与分布式训练等手段，实现高效扩展部署与监控 —— 将扩散模型驱动的功能部署上线，并持续监控模型漂移、延迟与输出质量跨职能合作 —— 与产品与设计团队协作，将视觉生成能力无缝融入用户体验中你是这样的人：喜欢独立拥有项目主导权相信“清晰来自行动” —— 擅长快速原型、测试与迭代，而非等待完美计划能在初创环境中保持冷静 —— 不惧优先级变化或从零开始构建系统做事迅速、执行力强 —— 相信“现在交付价值”比“完美但延迟交付”更重要将反馈与失败视为成长机会 —— 渴望不断精进自己拥有谦逊、求知欲与实干精神，并愿意带动他人成长任职要求在扩散模型与视觉生成领域具备扎实经验（如 Stable Diffusion、SDXL、Flux 等）熟练掌握 DreamBooth、LoRA/QLoRA 等微调技术精通 PyTorch（优先）具备图像数据准备能力，包括自动描述生成、标签标注、过滤与增强等熟悉 GPU 优化、潜变量扩散技巧与高效训练方法具备扎实的软件工程基础，包括算法、数据结构与整洁代码规范加分项有风格迁移、个性化生成或多模态生成方面的经验熟悉生成式视觉模型的安全对齐，如 NSFW 检测与对抗测试有开源扩散模型或视觉生成项目的贡献经验你将获得扁平化结构与真实的项目主导权全面参与产品方向与共识式决策弹性办公制度高度影响力岗位，与产品、数据与工程团队深度协作顶尖市场薪酬与绩效奖金全球化视野的产品开发机会丰厚福利 —— 住房租赁补贴、高质量公司餐厅、加班晚餐补助健康、牙科与视力保险全球旅行保险（适用于你和家属）无限制、灵活的带薪休假制度我们的团队与文化我们是一支高密度、高绩效的团队，专注于打造高质量产品，影响全球用户。我们像主人一样思考与行动，重视执行速度、沟通清晰与极致责任感。如果你渴望成长、追求卓越，欢迎加入我们！关于 BJAK BJAK 是东南亚最大的保险聚合平台，拥有超过 800 万用户，且公司完全由员工持股。总部位于马来西亚，业务遍及泰国、台湾与日本。我们通过 Bjak.com，帮助数百万用户获得透明、可负担的金融保障。我们通过 API、自动化与 AI 等前沿技术，简化复杂的金融产品，打造下一代智能金融系统。如果你对构建真实世界落地的 AI 系统充满热情，并希望在高影响力环境中快速成长，我们期待你的加入！ --------------------------------------------------------- Transform Visual Models into Real-World Applications We’re building AI systems for a global audience. We are living in an era of AI transition - this new project team will be focusing on building applications to enable more real world impact and highest usage for the world. This role is a global role with hybrid work arrangement - combining flexible remote work with in-office collaboration at our HQ. You’ll work closely with regional teams across product, engineering, operations, infrastructure and data to build and scale impactful AI solutions. Why This Role Matters You’ll fine-tune state-of-the-art models, design evaluation frameworks, and bring AI features into production. Your work ensures our models are not only intelligent, but also safe, trustworthy, and impactful at scale. What You’ll Do Fine-tune & Adapt – Train and customize diffusion models (SDXL, Flux, Stable Diffusion variants) using LoRA, DreamBooth, and other parameter-efficient methods. Curate Datasets – Build, clean, and annotate large-scale image datasets with captioning, tagging, and NSFW filtering for safe and aligned generation. Evaluate & Align – Develop pipelines to measure fidelity, diversity, style adherence, and safety across generated outputs. Optimize Performance – Apply GPU memory optimization, latent diffusion tricks, and distributed training for efficient scaling. Deploy & Monitor – Ship diffusion-powered features into production with monitoring for drift, latency, and quality. Collaborate & Deliver – Work with product and design to integrate generative vision capabilities into user experiences. What Is It Like Likes ownership and independence Believe clarity comes from action - prototype, test, and iterate without waiting for perfect plans. Stay calm and effective in startup chaos - shifting priorities and building from zero doesn’t faze you. Bias for speed - you believe it’s better to deliver something valuable now than a perfect version much later. See feedback and failure as part of growth - you’re here to level up. Possess humility, hunger, and hustle, and lift others up as you go. Requirements Strong experience with diffusion models and generative vision (Stable Diffusion, SDXL, Flux, etc.). Hands-on skills with DreamBooth, LoRA/QLoRA, and fine-tuning methods. Proficiency with PyTorch (preferred). Experience in dataset preparation (captioning, tagging, filtering, augmentation). Knowledge of GPU optimization, latent diffusion, and efficient training techniques. Strong foundations in software engineering, algorithms, and clean code practices. Nice to Have Prior work in style transfer, personalization, or multimodal generation. Experience with safety alignment in generative vision (NSFW detection, adversarial testing). Open-source contributions in diffusion or generative vision projects. What You’ll Get Flat structure & real ownership Full involvement in direction and consensus decision making Flexibility in work arrangement High-impact role with visibility across product, data, and engineering Top-of-market compensation and performance-based bonuses Global exposure to product development Lots of perks - housing rental subsidies, a quality company cafeteria, and overtime meals Health, dental & vision insurance Global travel insurance (for you & your dependents) Unlimited, flexible time off Our Team & Culture We’re a densed, high-performance team focused on high quality work and global impact. We behave like owners. We value speed, clarity, and relentless ownership. If you’re hungry to grow and care deeply about excellence, join us. About Bjak BJAK is Southeast Asia’s #1 insurance aggregator with 8M+ users, fully owned by its employees. Headquartered in Malaysia and operating in Thailand, Taiwan, and Japan, we help millions of users access transparent and affordable financial protection through Bjak.com. We simplify complex financial products through cutting-edge technologies, including APIs, automation, and AI, to build the next generation of intelligent financial systems. If you're excited to build real-world AI systems and grow fast in a high-impact environment, we’d love to hear from you.

Responsibilities

You will fine-tune state-of-the-art diffusion models and design evaluation frameworks to ensure their safe and trustworthy deployment. Your work will involve collaborating with cross-functional teams to integrate generative vision capabilities into user experiences.