SPEAKING · ENGINEERING EXCHANGE

Speaking, engineering & contact

I translate frontier AI into architecture decisions, working systems, and engineering lessons teams can act on.

AI Platform & Agentic AI Engineer · Agent systems · RAG · Evaluation

Send an invitation Connect on LinkedIn

PUBLIC SESSIONS

Selected talks

Public sessions backed by implementation notes, system artifacts, and architecture articles.

  1. COSCUP

    GenAI Workflow: Building an Intelligent Technology Trend Insight System

    Open-source data pipelines, GenAI workflows, and interactive visualization for continuously tracking technology trends.

    Explore the TrendScope implementation →
  2. iThome Cloud Summit

    Generative AI Platform Engineering for Financial Services: Building a Scalable Agentic AI Hub

    Platform engineering, MCP integration, multi-agent infrastructure, realtime inference, and cloud governance.

    Read the architecture article →
  3. iThome AI Enterprise Summit

    Financial-Grade Enterprise Agentic AI Architecture: From PoC to an Agentic Operating System

    Agentic RAG, multi-agent orchestration, self-verification, and LLM-as-a-Judge for high-stakes AI systems.

    Read the governance article →

SPEAKING TOPICS

Topics I can speak about

  1. From Agent Demo to Agent System

    Orchestration, tools, memory, evaluation, observability, permissions, and safety boundaries.

  2. RAG Beyond Retrieval

    Knowledge architecture, GraphRAG, multimodal documents, traceability, and evaluation.

  3. AI Engineering & Agent Harnesses

    Context engineering, long-running workflows, regression testing, and developer tooling.

SPEAKER BIO

Short speaker bio

Justin is an AI Platform and Agentic AI engineer who researches emerging models, agent systems, RAG, evaluation, and the engineering infrastructure required to move AI from demos into operable systems.

ENGINEERING PATH

Engineering background

My work connects document intelligence, retrieval, agent orchestration, evaluation, and platform operations.

  • Built medical-receipt OCR pipelines with preprocessing, layout detection, Traditional Chinese OCR optimization, table parsing, and multi-format normalization.
  • Implemented LLM-as-a-Judge evaluation, Agentic RAG, M365 knowledge assistance, realtime voice interaction, and AIGC workflows.
  • Currently focused on agent architecture, MCP, multi-agent coordination, evaluation, observability, and enterprise AI platform design.

EDITORIAL METHOD

Editorial method

Bloss0m is an engineering publication: AI can assist the workflow, but evidence, interpretation, and publication decisions remain accountable to the author.

  1. Primary evidence first

    Technical claims are traced to papers, official documentation, specifications, repositories, or first-party announcements whenever possible.

  2. Synthesis over summary

    Each article must add architecture, trade-offs, limitations, experiments, or an actionable engineering decision beyond its sources.

  3. Transparent AI assistance

    AI may support research organization, drafting, localization, and visual ideation. The author verifies material claims, links, bilingual parity, and final editorial judgment.

  4. Living technical notes

    Substantive corrections and re-verification update the modified date. Durable guides are maintained as topic hubs rather than republished as artificial freshness.

CONTACT

Useful details for an invitation

When possible, include the audience, topic, format, location or online setup, and expected date. I welcome conferences, internal engineering sessions, podcasts, and architecture exchanges.

[email protected] LinkedIn GitHub
Speaking · Exchange

Bring frontier AI and engineering reality into the same conversation.

For speaking invitations or technical exchange, send the context directly by email or LinkedIn.

Send an invitation