Agentic AI for Healthcare: Building Scalable Models for Production
暫譯: 醫療保健中的代理式人工智慧：構建可擴展的生產模型

醫療保健行業正在經歷一場變革，隨著人工智慧從孤立的預測模型轉向自主的、以目標為驅動的代理系統，這些系統能夠進行推理、規劃並實時適應。

《醫療保健的代理人工智慧：構建可擴展的生產模型》是第一本專門針對在複雜的臨床和運營醫療環境中設計和部署代理人工智慧系統的綜合技術指南。與傳統的人工智慧系統被動推斷結果不同，代理人工智慧涉及主動做出決策、與其他代理協調並朝著長期目標主動行動的系統，使其非常適合動態和高風險的醫療環境。本書的特點在於其深度的技術焦點，專為構建現實世界醫療解決方案的計算機科學家、架構師和人工智慧工程師量身定制。每一章都將理論基礎與使用雲原生技術、分佈式架構、知識表示模型和安全部署管道的生產就緒實現相結合。通過在腫瘤學、疫苗管理和醫院運營方面的詳細案例研究，本書展示了代理人工智慧如何改善護理交付、減輕臨床醫生的認知負擔並自動化常規工作流程，同時滿足監管標準。

隨著醫療保健越來越多地採用以人工智慧為首的策略，本書通過提供一種結構化的、以工程為中心的方法來構建可擴展的、智能的和可信的醫療系統，填補了這一關鍵需求。

Sairohith Thummarakoti is a technology leader who builds agentic AI systems that work in the real world. For more than a decade, he has delivered large-scale solutions across healthcare, health insurance, banking, finance, and fintech, using Pega Cloud to orchestrate planners, trusted tools, memory with provenance, and rigorous supervision. His focus is practical: designing cloud-native architectures, streamlining enterprise workflows, and embedding intelligence into mission-critical systems while maintaining security, performance, and cost efficiency.

In the Pega ecosystem, he's a Featured Member (top 52 among 764,000+ members) and a Top Contributor with over 500 accepted solutions spanning AI integration, decisioning, advanced design patterns, high availability, Pega Cloud 3 migrations, and secure application programming interface (API) integrations.

His agentic work centers on goal-pursuing software that perceives context, plans multi-step workflows, acts with approvals, explains decisions, and learns without drifting - always under governance. He has translated the agentic pattern - planner → tools → memory → supervision - into production services for healthcare and financial services. He treats autonomy as a dial, set deliberately to the lowest level that achieves the goal and only raised with evidence, guardrails, and human oversight. In practice, his systems have improved throughput and productivity, reduced operational cost, and tightened reliability in live applications. His deployment playbook includes equity and safety surveillance, uncertainty gating, typed tool contracts, and an incident response that is capable of pausing automation instantly.

Beyond delivery, Sairohith contributes to the global research and education community. He has authored IEEE conference papers, delivered invited talks, and presented at international forums on AI and cloud computing. As the founding Chair of the IEEE Computer Society - Columbia Section, he has organized conferences and workshops and delivered multiple Faculty Development Programs on agentic AI, helping faculty and students turn planner-tool stacks, human-in-the-loop (HITL) governance, and evaluation beyond accuracy into labs and capstone projects. He has worked as a book editor and authored several books on agentic AI, generative AI, and cloud infrastructure. He also mentors teams transitioning from prototypes to governed services. His throughline is consistent: Build systems that are explainable, auditable, and interruptible so they earn adoption where it matters most.

Sairohith Thummarakoti 是一位技術領導者，專注於構建能在現實世界中運作的自主人工智慧系統。十多年來，他在醫療保健、健康保險、銀行、金融及金融科技等領域提供大規模解決方案，利用 Pega Cloud 協調規劃者、可信工具、具有來源的記憶以及嚴格的監督。他的重點是實用的：設計雲原生架構、簡化企業工作流程，並將智能嵌入到關鍵任務系統中，同時保持安全性、性能和成本效率。

在 Pega 生態系統中，他是特色成員（764,000+ 成員中排名前 52 名）和頂尖貢獻者，擁有超過 500 個被接受的解決方案，涵蓋 AI 整合、決策、高級設計模式、高可用性、Pega Cloud 3 遷移以及安全的應用程式介面（API）整合。

他的自主工作集中於追求目標的軟體，能夠感知上下文、規劃多步驟工作流程、在獲得批准後行動、解釋決策並在不偏離的情況下學習——始終在治理之下。他將自主模式——規劃者 → 工具 → 記憶 → 監督——轉化為醫療保健和金融服務的生產服務。他將自主視為一個旋鈕，故意設置在達成目標所需的最低水平，並僅在有證據、護欄和人類監督的情況下提高。在實踐中，他的系統提高了產出和生產力，降低了運營成本，並在實時應用中增強了可靠性。他的部署手冊包括公平性和安全監控、不確定性閘道、類型化工具合約，以及能夠立即暫停自動化的事件響應。

除了交付，Sairohith 還為全球研究和教育社群做出貢獻。他撰寫了 IEEE 會議論文，發表了受邀演講，並在國際論壇上介紹人工智慧和雲計算。作為 IEEE 計算機學會哥倫比亞分會的創始主席，他組織了會議和研討會，並提供了多個關於自主人工智慧的教師發展計劃，幫助教師和學生將規劃者-工具堆疊、人機協作（HITL）治理以及超越準確性的評估轉化為實驗室和畢業專案。他曾擔任書籍編輯，並撰寫了幾本關於自主人工智慧、生成式人工智慧和雲基礎設施的書籍。他還指導團隊從原型轉型為受治理的服務。他的一貫主題是：構建可解釋、可審計和可中斷的系統，以便在最重要的地方獲得採用。