DennySORA
Search the site
Language：English

Language

name
李汶道
alias
DennySORA
role
Backend, cloud and AI systems engineering
status
Tokyo · studying Japanese
github
github.com/DennySORA（external site）
email
dennysora.main@gmail.com

DennySORA

I care about how software and AI systems move from an idea to something that runs, can be observed and can be maintained. My work has covered backend services, cloud platforms and agent integration; I am currently studying Japanese in Tokyo while continuing hands-on work with AI and LLMs.

What I work on

Professional practice, personal projects and what I am still learning are kept apart, so each ability has context and something you can check.

Backend & distributed systemsProfessional practice

From APIs and event processing to modular service boundaries, with attention to asynchronous flows, failure handling and how a system stays maintainable over time.

  • A plugin-based API platform (Trend Micro)
  • Kafka-based consumer services and a low-latency Go streaming server (KKStream)

ToolsGo · Python · Rust · Kafka · PostgreSQL · Redis

Cloud platforms & observabilityProfessional practice

Solving real deployment, resource-allocation and infrastructure problems with AWS, Kubernetes, Terraform and observability tooling.

  • Re-architected EKS node groups and moved monitoring to ARM nodes (Trend Micro)
  • Modular Helm charts and CNI troubleshooting (Trend Micro)

ToolsAWS · Kubernetes (EKS) · Terraform · Helm · OpenTelemetry · Prometheus

AI systems & personal model workProfessional practicePersonal projectsLearning

At work I built agent orchestration and streaming integrations; my personal research focuses on a trilingual model, inference and quantization, and I am still learning model interpretability.

  • ADK ↔ AGUI Middleware: SSE streaming and Human-in-the-Loop (Trend Micro)
  • Training a Chinese / English / Japanese decoder-only model from scratch (personal research)

ToolsGoogle ADK · LangGraph · vLLM · TensorRT-LLM · SentencePiece

Along the way

Roles and main work first; details for anyone who wants to go deeper. Based on the résumé already published on this site.

  1. — PresentEducation · current status

    Japanese Language Study，Tokyo School of the Japanese Language (Naganuma School)

    Full-time Japanese language study in Tokyo.

  2. —

    Senior Cloud Engineer，TrendMicro 趨勢科技

    Worked across agent orchestration, platform and API systems, cloud infrastructure and data protection.

    • Designed and shipped an agent-orchestration platform on Google ADK with clear app architecture and module boundaries.
    • Authored ADK ↔ AGUI Middleware (Python 3.13+) providing SSE streaming and Human-in-the-Loop workflows; integrated with the orchestrator.
    • Re-architected EKS node groups: moved monitoring to ARM nodes for cost savings and introduced dedicated CPU-/memory-optimized pools.
    Show all 24 work items

    Companion-Orchestrator Agent Center

    • Designed and shipped an agent-orchestration platform on Google ADK with clear app architecture and module boundaries.
    • Implemented Agent-to-Agent (A2A) communication to coordinate remote agents.
    • Built an LLM agent with tool use, state tracking, retries, and error handling.
    • Delivered a web Playground for scenario testing and prompt/tool debugging; frontend developed independently with AI-assisted workflows.
    • Created a CopilotKit adapter and composed multi-step LangGraph flows.
    • Enabled dynamic agent loading via database-backed configs for zero-downtime rollout.
    • Authored ADK ↔ AGUI Middleware (Python 3.13+) providing SSE streaming and Human-in-the-Loop workflows; integrated with the orchestrator.
    • Achieved an end-to-end pipeline A2A → ADK → LangGraph → CopilotKit adopted cross-org with other teams.

    Platform Development & API Management

    • Built a plugin-based API platform that lets teams ship integrations without core changes; provided lifecycle hooks, auth, and rate limiting.
    • Developed a Collector Manager for configurable log routing with backpressure and retry controls.
    • Created a Mini-Companion system with GitHub-based sync for declarative workflows.
    • Implemented a Companion Proxy enforcing PII redaction and policy-based request filtering.

    Cloud Infrastructure Optimization

    • Re-architected EKS node groups: moved monitoring to ARM nodes for cost savings and introduced dedicated CPU-/memory-optimized pools.
    • Resolved Kubernetes compatibility defects in Stream Lake environments.
    • Authored modular Helm charts to standardize deployments across services.
    • Fixed CNI IP exhaustion and routing issues to stabilize multi-tenant clusters.

    AI Integration & Data Security

    • Built an AWS Comprehend PII pipeline that reduced costs by ~99% while preserving detection quality.
    • Created an AI Proxy Adapter with Amazon Bedrock integration and policy enforcement.
    • Developed a Kafka → Neptune pipeline for MITRE-aligned knowledge graphs.
    • Implemented fine-grained AI access controls with auditability and tenant isolation.

    Technical Leadership

    • Produced executive-level AI demos for the CEO's NVIDIA presentation.
    • Led internal workshops on AI tooling and best practices.
    • Recognized as Employee of the Quarter.
    • Drove cross-functional enablement through reusable platform components.
    ADK-AGUI Middleware (GitHub)（external site）

  3. —

    Back-End Engineer，KKStream

    Worked on streaming services, a Kafka consumer architecture and CI/CD, supporting deployments with infrastructure as code.

    • Designed Kafka-based consumer service architecture.
    • Built low-latency streaming server using Golang.
    • Implemented Terraform for infrastructure as code.
    Show all 12 work items

    DevOps & Infrastructure

    • Optimized GitLab CI/CD pipelines, reducing deployment time by 40%.
    • Deployed Kubernetes clusters using Helm.
    • Implemented Terraform for infrastructure as code.
    • Managed Kubernetes clusters using Kops.

    System Architecture & Development

    • Designed Kafka-based consumer service architecture.
    • Created Docker Compose development environment.
    • Integrated multiple backend services and databases.
    • Built low-latency streaming server using Golang.

    Research & Implementation

    • Evaluated WebRTC Mesh and SFU technologies.
    • Integrated Pydantic for robust data validation.
    • Refactored monolithic projects for better maintenance.
    • Implemented CI/CD pipeline testing integration.

  4. —

    Back-End Engineer (Intern)，天堂遊戲有限公司

    Developed and profiled RESTful APIs in Go, and took part in data processing and technology evaluation.

    • Optimized RESTful APIs using Golang PProf.
    • Implemented Apache Spark for data processing.
    Show all 6 work items

    API Development & Optimization

    • Optimized RESTful APIs using Golang PProf.
    • Reduced API response times by 30%.
    • Refactored codebases for improved modularity.

    Research & Evaluation

    • Implemented Apache Spark for data processing.
    • Demonstrated GraphQL API capabilities.
    • Evaluated emerging technologies for adoption.