Cake Job Search

開發與實作 生成式 AI 與影像辨識模型，用於產品瑕疵生成（Stable Diffusion、ControlNet、GAN 及其變體）、資料擴增與品質檢測。應用與微調 影像辨識／表徵學習模型（如 YOLO、CLIP、DINOv3 等新型開源模型）。進行 大型模型（ViT、LLM）Fine-tuning、模型建置、調校與效能優化。建立 資料清理、特徵工程、模型訓練與評估流程，持續提升模型穩定性與準確度。規劃並實作 RAG（檢索增強生成）架構，整合向量資料庫與 LLM。開發與維護 AI 模型服務 API，並與 WPF 前端及其他系統整合。進行 AI 技術與學術文獻研究，將研究成果應用於實務專案。使用 Docker、Git 進行部署與版本控管，配合主管交辦事項與團隊合作。依專案需求配合國外出差（一年約 1--3 次，每次約 3--4 週），參與專案導入、系統部署、模型調校或跨部門協作。其他主管交辦事項，並與團隊密切合作完成專案目標。必備條件3 年以上 AI／機器學習／深度學習 相關工作經驗，熟悉python、C#。熟悉 TensorFlow、PyTorch、Keras 等深度學習框架。熟悉 影像辨識模型與演算法，具 YOLO、CLIP、DINO（含新版本）等實作或應用經驗。具備 生成式模型（GAN 及其變體、Stable Diffusion、ControlNet）實務經驗，並理解模型底層原理。熟悉 Hugging Face 生態系（Transformers、PEFT、Datasets、Tokenizers）。有使用 PEFT 技術經驗（如 LoRA、QLoRA、Prefix Tuning 等），能在有限 GPU 資源下進行有效訓練與調校。了解 Fine-tuning 對模型效能、泛化能力與推論成本的影響，並能依量產需求進行權衡與最佳化。具備 RAG 架構、Prompt Engineering、向量搜尋與任一向量資料庫 經驗。具備 API 開發經驗，熟悉 Docker、Git。具備 數據清理、特徵工程、模型評估與效能優化能力。具良好溝通能力，可獨立作業並進行團隊合作。加分條件曾參與 產品瑕疵檢測、工業影像或製造業 AI 專案。具 影像生成＋影像辨識混合應用 經驗（如生成資料輔助訓練）。具 論文閱讀與模型重現（Reproduction）經驗。具 WPF 前端開發 或 AI 系統整合經驗。曾參與 AI 相關研究、技術分享或開源專案。具 對比學習（Contrastive Learning） 應用經驗（如 Self-Supervised Learning、Representation Learning）。有使用 StableDiffusion WebUI、ComfyUI經驗。

公司介紹 這是一家高速成長的 AI 新創團隊，專注於 Physical AI 與 Industrial AI 的落地應用。團隊致力於打造能將真實世界工業環境快速轉換為高擬真模擬場景的 AI 平台，讓 AI 能在虛擬環境中進行訓練與策略優化，最終部署至真實工業系統。 公司產品結合 3D Simulation、Digital Twin 與 Reinforcement Learning 技術，讓企業可以在虛擬世界完成複雜系統的訓練與測試，大幅降低現實環境的成本與風險，並加速智慧製造與自動化的導入。 團隊由多位具創業與深度技術背景的成員組成，目前已獲得 國際級 AI 科技公司與知名投資機構支持，並與高端製造及半導體產業合作。工程師將有機會參與 AI + Simulation + Robotics 的前沿技術應用，直接將研究級技術落地至真實世界場景。 工作內容 作為 Senior Reinforcement Learning Engineer，你將負責透過強化學習技術優化自動化系統與工業場景中的 AI 決策能力，並在模擬環境中訓練可部署於真實世界的 AI 模型。 你將有機會參與 AI 在 工業自動化、機器人與智慧製造場景中的實際應用。 主要職責包含： 在 NVIDIA Isaac Sim 等模擬平台上建立與訓練強化學習模型 設計並實作 state space、action space 與 reward function 開發與優化 RL policy（如 Q-learning / Policy-based methods） 與 AI / Simulation / 3D 團隊合作進行系統模擬與模型訓練 持續調整模型與參數，提升策略效率與系統穩定度 將模擬環境中的最佳策略導入實際工業系統 使用技術 Python PyTorch / TensorFlow Reinforcement Learning Algorithms NVIDIA Isaac Sim Simulation / Digital Twin Robotics / Automation Systems

公司介紹 我們的客戶是一家專注於 Physical AI 與數位孿生（Digital Twin）技術 的 AI 新創團隊，致力於打造能夠讓 AI 在虛擬環境中訓練、學習並優化決策的下一代平台，並將最佳策略部署回真實世界場域。 團隊核心技術結合 AI、3D 模擬、運籌優化與大型場域建模，應用於智慧製造、自動化物流與高精密產業等場景，協助企業在虛擬工廠中完成策略模擬與 AI 訓練，大幅提升建置效率與營運決策能力。 公司獲得 國際 GPU 生態系與多家產業資本支持，並與大型科技與製造企業合作，是少數專注於 Sim-to-Real / Real-to-Sim AI 應用落地 的技術團隊。 如果你對 AI + Simulation + Robotics / Automation 的跨領域技術有興趣，這將是一個能直接參與核心產品與演算法設計的機會。 工作內容 你將加入核心演算法研發團隊，負責設計並優化應用於 大型數位孿生場域的演算法模型，協助解決複雜環境中的路徑規劃、排程與資源配置問題。 主要職責包含： 建立與設計演算法模型（如圖論、網格建模、路徑規劃或優化模型） 從 3D 環境與場域資料中建構抽象模型並設計解法 設計並實作高效能演算法模組（Python 為主） 使用 GPU 加速工具優化大型資料處理或模擬效能 與產品與工程團隊合作，將演算法整合至平台系統 測試與評估不同策略表現並持續優化 使用技術 Python CUDA / GPU Parallel Computing PyTorch Algorithm Design / Data Structures Graph Theory / Path Planning 3D Simulation / Digital Twin Optimization / Scheduling

Logitech is the Sweet Spot for people who want their actions to have a positive global impact while having the flexibility to do it in their own way.The Role: Interns at Logitech work on a variety of tasks.  Interns will be responsible for understanding customer needs and developing solutions that meet their needs. You will also work with other teams to define strategies and drive outcomes.  You will receive support from their teams and have opportunities to grow in knowledge of your specified area.  Your Contribution: Your Qualifications and Skills: - Data analytics experience, and ability to work with Excel to explore and analyze data - Proficient in Google suite  - Proficiency creating reports and presentations  - Comfortable leading a discussion with a group of people - Ability to create reports and presentations  - Comfortable thinking out of the box and developing new ideas - Systems thinking able to synthesize disparate ideas into conclusions - Ability to convert data into insights - Presentation skills and ability to tell a story with data - Our ideal candidate is motivated, organized, and has a passion for technology and product management - Strong problem-solving skills, excellent communication skills, and the ability to work in a fast-paced environment Your Contribution:Be Yourself. Be Open. Stay Hungry and Humble. Collaborate. Challenge. Decide and just Do. These are the behaviors you’ll need for success at Logitech. In this role you will:Algorithm Implementation: Implement image signal computation algorithms to evaluate image quality and system performance.Machine Learning Application: Apply Machine Learning/Deep Learning (ML/DL) models to practical image processing tasks (e.g., color, brightness, denoising, enhancement, or scene recognition).Architecture Integration: Work within the Android Camera Framework to ensure seamless and efficient data flow between the Image Signal Processor (ISP) and the application layer..Research Design: Conduct and integrate objective experiments, and take charge of designing and analyzing subjective testing scenarios.Key Qualifications:We are looking for a proactive learner with a solid foundation in mathematics, familiarity with Android internals, and the ability to use development tools efficiently.Educational Background: Currently pursuing a Bachelor’s or Master’s degree in Computer Science, Electrical Engineering, Imaging Science, or a related field.Core Technical Skills:Proficiency in C/C and Python programming.Strong knowledge of Digital Image Processing (DIP) and Signal Computation (e.g., matrix operations, frequency domain analysis).Practical experience with Machine Learning/Deep Learning frameworks (e.g., PyTorch, TensorFlow) for computer vision problems.Domain Expertise:Familiarity with Android Architecture: Solid understanding of the Android Camera stack (HAL/Framework) and image data handling processes.Proficiency in Git: Experience with version control, including branch management, merging, and resolving code conflicts.Understanding of Image Quality (IQ) metrics and experimental methodologies.Soft Skills:Strong analytical and problem-solving abilities.Excellent communication and teamwork skills to collaborate across software and hardware domains.English Proficiency: Good written and verbal English communication skills for a global working environment.#LI-KS1/104Across Logitech we empower collaboration and foster play. We help teams collaborate/learn from anywhere, without compromising on productivity or continuity so it should be no surprise that most of our jobs are open to work from home from most locations. Our hybrid work model allows some employees to work remotely while others work on-premises. Within this structure, you may have teams or departments split between working remotely and working in-house.Logitech is an amazing place to work because it is full of authentic people who are inclusive by nature as well as by design. Being a global company, we value our diversity and celebrate all our differences. Don’t meet every single requirement? Not a problem. If you feel you are the right candidate for the opportunity, we strongly recommend that you apply. We want to meet you!We offer comprehensive and competitive benefits packages and working environments that are designed to be flexible and help you to care for yourself and your loved ones, now and in the future. We believe that good health means more than getting medical care when you need it. Logitech supports a culture that encourages individuals to achieve good physical, financial, emotional, intellectual and social wellbeing so we all can create, achieve and enjoy more and support our families. We can’t wait to tell you more about them being that there are too many to list here and they vary based on location.All qualified applicants will receive consideration for employment without regard to race, sex, age, color, religion, sexual orientation, gender identity, national origin, protected veteran status, or on the basis of disability.If you require an accommodation to complete any part of the application process, are limited in the ability, are unable to access or use this online application process and need an alternative method for applying, you may contact us toll free at 1-510-713-4866 for assistance and we will get back to you as soon as possible.