HOMEE AI_睿締國際科技股份有限公司

About the Team The Spatial Intelligence CV group builds production-grade computer vision systems that power multimodal and spatially-aware products, from property imagery understanding to 3D scene analysis. We focus on delivering reliable, high-performance services—turning state-of-the-art vision models into scalable, user-facing products. About the Role You will design, optimize, and scale computer vision inference software running on both on-prem and cloud GPU clusters. Beyond writing production-quality Python, you will integrate advanced vision models into end-to-end pipelines that deliver real business value in domains such as architecture, real estate, and interior design. Your work will directly improve system accuracy, throughput, and cost efficiency. In this role, you will Build reusable Python inference pipelines for object detection, segmentation, and image generation at scale. Implement and optimize training/inference flows for YOLO, U-Net, Segment Anything Model (SAM), and Stable Diffusion. Fine-tune diffusion and LoRA-based models for domain-specific generation and augmentation. Develop and integrate solutions in Novel View Synthesis, 6D Pose Estimation, Monocular Depth Estimation, Visual Question Answering (VQA), and Style Transfer. Build Scene Graph pipelines to support space design, dimension measurement, and spatial understanding. Conduct performance profiling and latency/throughput tuning for CV back-ends on GPU clusters. Collaborate with product teams to deliver features such as automated property feature extraction, content tagging, and high-quality image synthesis. You might thrive in this role if you Enjoy turning cutting-edge CV and spatial reasoning research into robust, scalable production services. Have experience across multiple vision tasks and can quickly adapt to new architectures. Are motivated by measurable impact—accuracy gains, inference speedups, or improved user experience. Communicate clearly with cross-functional stakeholders in Mandarin. Minimum Qualifications Master’s degree (or higher) in CS, EE, Applied Math, or related field, or equivalent practical experience. One or more academic or professional projects related to computer vision. Hands-on experience with OpenCV, PyTorch, Git, Docker. Practical experience with YOLO, U-Net, Segment Anything, and Stable Diffusion. Familiarity with at least two of the following: Novel View Synthesis, 6D Pose Estimation, Monocular Depth Estimation, VQA, Style Transfer. Understanding of Scene Graph concepts and their application in spatial reasoning tasks. Bonus Qualifications Experience in classification, segmentation, detection across various datasets. Familiarity with ControlNet or other controllable generation techniques. Publications in top-tier CV/AI conferences (CVPR, ICCV, ECCV, NeurIPS, ICML). Experience operating large-scale CV inference services with measurable performance metrics.

About the Team The Spatial Intelligence LLM group builds production-grade inference stacks that power next-generation conversational and multimodal experiences grounded in text and spatial data (imagery, maps, 3-D scenes). We focus on shipping reliable, high-performance services—turning proven techniques into robust, scalable products. About the Role You will design, optimize, and scale LLM inference software that runs on both on-prem and cloud GPU clusters. Beyond writing production-quality Python, you will work with platform engineers to integrate techniques. Your contributions will directly improve system latency, throughput, and cost efficiency. In this role, you will Build reusable Python inference pipelines serving chat, RAG, and VLM workloads at scale. Conduct performance profiling and latency/throughput tuning for LLM back-ends (vLLM, TensorRT-LLM, Triton). Track the latest developments in model architecture (e.g., Phi-3, Gemma, Qwen 2.5) and rapidly prototype optimizations with clear product impact. Collaborate with product teams to deliver features such as hallucination mitigation, adaptive RAG pipelines, and agent services. Provide feedback to infrastructure teams on GPU utilization, memory bandwidth, and batching strategies. You might thrive in this role if you Enjoy transforming ideas into production code that ships on a regular cadence. Value clean abstractions (PEFT, TRL, vLLM) that enable fast experimentation. Are motivated by measurable impact—milliseconds saved, tokens /sec gained, costs reduced. Communicate clearly with cross-functional stakeholders in Mandarin. Minimum Qualifications Master’s degree (or higher) in CS, EE, Applied Math, or equivalent practical experience. 3+ years building and deploying deep-learning systems in Python. Proficiency with PyTorch and the Hugging Face ecosystem; solid software-engineering practices (tests, CI/CD, Docker). Demonstrated knowledge of LLM internals (attention, KV caching, quantization) and modern inference frameworks (vLLM, TensorRT-LLM, Triton). Bonus Qualifications Publications or technical blogs on practical LLM deployment or optimization. Experience operating large-scale production ML services with SLO/SLI ownership.

HOMEE AI （https://homee.ai）創造了空間設計生成式 AI，幫助使用者輕鬆打造室內空間。HOMEE 合宜家居 (https://homee.cc/) 是 HOMEE AI 在台灣創造的試驗場域，應用 HOMEE AI 獨家開發的空間技術，協助全球家居產業數位轉型。 我們的團隊充滿了熱情和專業，來自不同領域的卓越人才匯聚在一起，克服各種挑戰。如果你擁有創新思維，懷抱讓人類生活更美好的夢想，我們非常歡迎你加入我們的團隊！ 身爲我們的財務夥伴，我們希望能擁有以下優秀的經驗與技能。 財務出納管理（含應計及應收款項等收付相關作業以及ERP帳務處理） 1.財務出納管理、應付及應收款等收付款相關作業(含ERP帳務處理)。 2.關係企業公司之相關資金管理。 3.銀行往來相關聯繫、帳款收支操作與匯款資料核對作業。 4.銀行額度申請資料準備及額度、授信合約控管維護。 5.財務報表等相關報表製作。 6.配合會計師查帳提供相關資料。 7.其他財會專案事宜。 8.主管交辦事項。