Stefanus Agus Haryono

Led backend development for Tandem, a generative AI platform serving 40,000+ healthcare professionals, contributing to frontend improvements and maintaining >99% availability. Owned the architecture and end-to-end development of SelfServeBot, a self-service RAG platform with LangChain/LangGraph and Azure AI services, reducing chatbot setup time from 2+ weeks to under 10 minutes and enabling 1,000+ users to launch 800+ document-grounded chatbots. Architected and developed the Tandem API, a secure LLM API with JWT authentication, rate limiting, and API management for 12 models processing 150,000+ prompts and ~1B tokens annually across 30+ applications. Built AgentSea for configurable agent creation with model selection, custom prompts, document skills, and Microsoft 365 integrations, including security features with Bedrock Guardrails, PII masking, and adversarial-input protection. Developed a generative AI-assisted clinical-trial matching system that screened 8,000+ patients with >90% a

Led development and maintenance of a deep learning-based medical ICD recommendation system used by physicians across 5+ hospitals, improving recall by up to 10%. Built backend services for Hospital Information Systems and AI applications supporting physician workflows. Owned feature delivery, release management, and production deployments within an Agile cycle. Managed AI service deployments on Azure and GCP, implementing CI/CD pipelines and production monitoring with Power BI and Azure Application Insights while mentoring junior engineers and interns.

Conducted research in software engineering and machine learning under Prof. David Lo, developing prototypes for software evolution, static analysis, API migration, vulnerability analysis, recommender systems, explainable AI, and transformer-based text classification.