Tìm kiếm việc làm online mới nhất!

Mở rộng tìm kiếm
Tắt
Neihu District, Taipei City, Taiwan
Full-time
Trung cấp
工作內容1. 負責公司 AI 產品中離線小型模型(On-device / Edge Model)的訓練、微調(Fine-tuning)與部署優化。2. 針對特定應用場景,對大型語言模型(LLM)或其他 AI 模型進行知識蒸餾(Knowledge Distillation),產出符合效能與精度需求的小型模型。3. 執行模型重新訓練(Retraining),根據新資料或需求變更持續迭代模型表現。4. 設計並執行模型微調策略,包含 LoRA、QLoRA、Adapter Tuning、Prompt Tuning 等方法,以最小成本達成模型客製化。5. 建立與維護模型訓練 Pipeline(資料前處理、訓練、評估、部署),確保流程可重現且自動化。6. 進行模型壓縮與優化(量化 Quantization、剪枝 Pruning、蒸餾 Distillation),使模型能在邊緣裝置或離線環境高效運行。7. 設計模型評估指標與基準測試(Benchmark),確保微調後模型品質符合產品需求。8. 與產品團隊、資料團隊及後端工程師協作,將訓練完成的模型整合至產品中。9. 追蹤 AI/ML 領域最新技術趨勢(如開源模型、訓練框架更新),持續引入最佳實踐。
Negotiable
Yêu cầu 3 năm kinh nghiệm
Không yêu cầu kinh nghiệm quản lý
[About Us 醫療X科技] 慧康生活科技 Health2Sync 成立於 2013 年,旗下擁有「智抗糖 App」與「雲端智慧照護平台」兩大產品,是同時具有B2B2C的數位新創團隊。主打產品「智抗糖App」為亞洲最大慢性病管理平台,提供個人化並可複製的慢性病管理方案,建立新一代的數位療法,這套數位醫療解決方案在全球累計已有超過百萬名使用者。我們的服務結合簡易操作的數位工具與個人化及自動化的即時分析,串連慢性病友的關懷照護網路與生活數據,協助糖友透過 App 自我管理血糖,並讓家人與醫療單位透過「雲端照護平台」更即時瞭解狀況並提供關懷。透過這套解決方案,我們積極串起糖尿病照護生態圈,連結藥廠與病患,讓藥廠提供個人化的衛教及介入 (engagement),進而提升用藥的依從度,讓病患能有較好控制效果。[ About Our Team ]在數位醫療領域中,數據團隊是不可或缺的角色。作為H2數據團隊的一員,你將有大量的機會接觸到多樣化的生醫數據,通過數據分析探索及機器學習技術,最終的數據產品將可以直接回饋給用戶或醫療院所,讓數據發揮最大的價值。另外,數據團隊也有機會參與公司的重大決策,你的分析成果將會影響產品的走向及各項商業發展。在我們團隊中每個人都有機會一起發想及創造全新的數據產品,同時每個人的意見在團隊中都是非常重要的。期待在數據科學領域有豐富經驗的你可以加入我們,一起學習如何用數據科學相關的技術,為病患及醫療人員開發更多有價值的數據產品。 [Job Description]- 推動數據科學專案,以優化H2 產品。- 利用數位健康數據開發並部署機器學習, 統計模型及生成式 AI (GenAI) 應用。- 針對產品及分析應用,構建並維護基於大語言模型 (LLM) 的應用。- 優化並維護現有的 AI 系統,確保其性能與可擴展性。- 建立並改進數據基礎設施(data pipeline, data lakehouse), 以支援數據分析與建模。- 透過分析工具與統計方法,支援產品及業務團隊任務。
Python
PostgreSQL
Machine Learning
1.1M+ TWD / năm
Yêu cầu 3 năm kinh nghiệm
Không yêu cầu kinh nghiệm quản lý
About the Role This is not a traditional, desk-bound software engineering role. We are looking for a highly autonomous Forward Deployed Engineer (FDE) to act as the technical bridge between our core AI products and our enterprise clients. You will be deployed directly to client sites globally to integrate our AI systems and necessary edge point into their existing, often legacy, local network infrastructures. We do not care what programming languages you have memorized. We care about your ability to act as an orchestrator of code—using AI tools to rapidly generate, test, and deploy functional logic across any environment. You must possess the architectural understanding to define the problem, the "vibe coding" skills to accelerate the solution, and the extreme ownership to see it running reliably in production. Core Responsibilities: Intent-Driven Development ("Vibe Coding"): Leverage AI coding agents (e.g., Cursor, GitHub Copilot, Claude) to rapidly generate boilerplate, write functional logic, and build API connectors. You will operate as a polyglot, adapting instantly to whatever language or framework the client's legacy environment demands.On-Site Systems Integration: Manage the physical and digital deployment of AI hardware and software systems at enterprise client sites. Map local networks, establish secure connections, and force disparate systems to communicate.Autonomous Problem Discovery Resolution: Do not wait for perfectly scoped Jira tickets. You will proactively interrogate client systems, identify latency, failure points, or network blockers, and independently architect and deploy the fix on the spot.AI Debugging Hallucination Management: AI generates the code, but you own the deployment. You must possess the deep systems knowledge required to instantly spot AI hallucinations, security flaws, or inefficient logic in the generated output before it impacts a live client environment.Overseas Deployment Client Management: Travel internationally a minimum of 30% of the time. You must navigate undocumented local networks and communicate technical realities directly to regional stakeholders without escalating every issue back to headquarters.
Vibe Coding
LLM
intergration
Negotiable
Yêu cầu 3 năm kinh nghiệm
Không yêu cầu kinh nghiệm quản lý
1. 開發生成式AI相關應用2.協助資料收集、清理與處理文本/音檔資料]3. 參與 LLM 與語音模型(如 Whisper) 的微調與優化4. 協助開發與測試模型、提升文本生成與語音辨識的準確性與相關性5. 研究並嘗試不同的模型整合與優化方法6.其他主管交辦事項*必要技能: 1. 熟悉機器學習、自然語言處理(NLP)、向量資料庫或 RAG 技術有基礎理解2. 熟悉 Python,並了解基本的機器學習/深度學習工具(如 PyTorch、TensorFlow、scikit-learn任一皆可)3. 對LLM或語音AI模型有興趣,願意投入研究與實作,具備良好學習能力及問題解決能力
60K ~ 90K TWD / tháng
Yêu cầu 3 năm kinh nghiệm
Không yêu cầu kinh nghiệm quản lý
設計與實作 LLM 微調(Fine-Tuning)流程,並根據實務應用需求進行模型優化運用 HuggingFace、LoRA、QLoRA、Deepspeed 等工具對 Transformer 架構進行調校建構與維護資料前處理與 Tokenization pipeline,支援中文與多語環境下的 NLP 任務撰寫並執行模型評估流程(BLEU / ROUGE / EM / Retrieval-based Precision 等)部署並持續優化已訓練模型於雲端環境中的表現與穩定性與產品開發團隊協作,提供技術建議與最佳實作,落地 AI 解決方案
70K ~ 100K TWD / tháng
Yêu cầu 3 năm kinh nghiệm
Không yêu cầu kinh nghiệm quản lý
Key Responsibilities:1.Design, develop, and maintain scalable and efficient backend services to support NLU workflows.2. Integrate and optimize vector databases for storing and querying embeddings.3. Develop APIs and services to expose NLU capabilities to other systems and applications.4. Ensure data pipelines and backend services meet performance, security, and scalability requirements.5. Research and implement best practices for integrating LLMs and vector databases into backend architectures.
Negotiable
Yêu cầu 3 năm kinh nghiệm
Không yêu cầu kinh nghiệm quản lý
建構與維護 AI 應用微服務架構,使用 FastAPI、Pydantic、SQLAlchemy 等技術堆疊整合 OpenAI API、HuggingFace 模型與內部 LLM 推論流程,開發彈性可擴展的 inference pipeline與資料科學家合作,協助 AI 模型部署與效能最佳化使用 Docker 與 K8s 建立容器化部署流程,實現 DevOps 自動化與多環境可遷移性實作 CI/CD pipeline(GitHub Actions、Terraform 等),佈署至 GCP/AWS 等雲端平台設計穩定與高可用的 RESTful API / event-driven 架構,提供高品質產品對外服務介面參與跨部門產品設計與開發週期,實現 AI 功能與用戶需求之間的最佳整合
70K ~ 100K TWD / tháng
Yêu cầu 3 năm kinh nghiệm
Không yêu cầu kinh nghiệm quản lý

Tìm kiếm việc làm Cake

Tham gia Cake ngay bây giờ! Tìm kiếm công việc hoàn hảo của bạn giữa hàng chục nghìn việc làm khác nhau.