Johnson Jao johnson00111

Hi there, I'm Johnson Jao 👋

Johnson Jao | Software Engineering | Machine Learning | Agentic AI

I am a Master of Telecommunications candidate and Course Assistant at the University of Pittsburgh. I leverage 3+ years of industry experience at Trend Micro to build AI systems that are not just accurate, but scalable, secure, and autonomous.

Focus: Agentic AI, LLM Orchestration, and MLOps.
Mission: Transitioning from building robust backends to architecting intelligent, self-evolving AI agents.
Academic Role: Course Assistant for the MDS program @ Pitt SCI.
Side Kick: I adopted a virtual Golden Retriever called Bob (@bobtheagent) — my loyal bot who ships vibe-coded projects for me. 🐕

Current Focus: Stateful Multi-Agent Orchestration

I am currently exploring autonomous, hierarchical corporate structures within AI systems.

ai-agent-office `In-Progress`

Building an MVP for a hierarchical multi-agent collaboration framework.

Ongoing Work: Experimental implementation of a Supervisor Orchestrator using LangGraph to manage complex states between specialized worker nodes.
Architectural Goal: Transitioning from linear chains to a scalable, role-playing organizational framework.

agentic-ai-learning `Ongoing Study`

A systematic knowledge base for agentic reasoning patterns.

Research Scope: Implementing and benchmarking ReAct, Self-Reflection, and Planning patterns.
Current Objective: Documenting orchestration strategies and tool-use capabilities to establish high-reliability agent standards.

Recent Projects

math-agent-eval `New`

Benchmarking LLM agent reasoning strategies on mathematical problem-solving.

Architecture: 4 LangGraph agents — Chain-of-Thought, ReAct, Constrained Reasoning, and Multi-Agent collaboration — evaluated across 30 math problems.
Observability: End-to-end tracing and scoring with Langfuse v4, enabling side-by-side comparison of agent accuracy and reasoning quality.

gmail-job-app-tracker `New`

An AI-powered pipeline that turns your inbox into a job search dashboard.

Pipeline: Gmail API → dual Ollama models for classification & extraction → SQLite storage → React dashboard.
Visualization: Interactive Sankey funnel for application stages, trend charts, and a built-in AI chat assistant for querying your job search data.