Applied AI Engineer focused on building production-oriented AI systems that automate real-world workflows.

I design and deploy multi-agent systems that move beyond prototypes into reliable, task-oriented tools.

Project Mirror is a multi-agent AI system designed to act as a professional assistant—handling tasks like information retrieval, scheduling, and technical reasoning through coordinated agent workflows.

👉 It demonstrates how AI systems can operate in real environments, where outputs directly influence decisions and actions.

Short walkthrough demonstrating multi-agent coordination, task execution, and real workflow automation.

• Multi-agent orchestration for complex task execution
• Retrieval-augmented reasoning (RAG) with strict context isolation
• Real-world tool integration (e.g., scheduling via APIs)
• Structured outputs and validation for reliability
• Observability into system behavior and failure modes

Most AI projects demonstrate isolated capabilities.
Project Mirror focuses on system reliability, coordination, and real usability.

Project Mirror employs a hierarchical orchestration pattern where a high-reasoning "Nexus" agent manages a fleet of specialized sub-agents.

graph TD
    User([User]) <--> Frontend[Next.js 16 / Tailwind CSS 4]
    Frontend <--> API[FastAPI Backend]

    subgraph "MAS Orchestration (Google ADK)"
        API <--> Nexus{Nexus Orchestrator<br/>Gemini 3.1 Flash-lite}
        Nexus -- "Control Handoff" --> DemoSpec[Demo Specialist<br/>Llama 3.3 70B]
        Nexus -- "Tool Call" --> Researcher[Researcher Agent<br/>Llama 3.3 70B]
        Nexus -- "Tool Call (MCP)" --> Calendar[Google Calendar / Meet]
    end

    subgraph "Knowledge & Tools"
        Researcher <--> VectorDB[(Weaviate Augmented DB)]
        Researcher <--> Search[Google Search API]
        DemoSpec <--> MockData[(Isolated Demo Contexts)]
        Calendar <--> GoogleAPI[(Google APIs)]
    end

    subgraph "Reliability & Privacy"
        Nexus -.-> RedTeam[Adversarial Red-Team]
        RedTeam -.-> Guardrails[Pydantic Validation]
        Guardrails -.-> Nexus
        API -.-> Fingerprint[SHA-256 Fingerprinting]
    end

Project overview and supporting documentation: https://github.com/Hou-dini/project-mirror-overview

Kognia AI is a hierarchical multi-agent system for autonomous research synthesis, strategic analysis, and report generation to support fast decision making.

• Automated research and analysis workflows using agent coordination
• Real-time orchestration visibility and logging
• Structured reasoning pipelines for consistency

➡️ Source Code

• Designing multi-agent systems that handle real tasks
• Improving LLM reliability through validation and grounding
• Building systems that balance latency, cost, and accuracy
• Turning complex workflows into usable AI tools

AI / Systems

• Multi-agent orchestration (Google ADK, MCP)
• RAG systems (Weaviate, structured outputs)
• Model routing and evaluation

Backend

• Python (FastAPI, asyncio)
• REST APIs, microservices

Data

• PostgreSQL, MongoDB
• Vector databases (Weaviate)

Infra

• Docker, CI/CD (GitHub Actions)
• Cloud deployment (GCP, Vercel)

• Email: elikplimkudowor@gmail.com
• LinkedIn: https://linkedin.com/in/elikplim-kudowor