I transform complex manual workflows into autonomous multi-agent systems. With a background in Software Engineering and Quantitative Analysis, I bridge the gap between high-level reasoning and robust, production-ready AI orchestration.
I don’t just build "AI wrappers." I architect Hierarchical Multi-Agent Systems that reason, plan, and execute. My core focus is solving the fragility of traditional LLM implementations by applying modular task decomposition, RAG-grounded execution, and adversarial risk-checking.
My Architectural Philosophy:
- Decoupled Reasoning: Utilizing the Nexus-Specialist pattern to mitigate LLM hallucinations.
- Grounded Execution: Ensuring every agentic decision is backed by verifiable, real-time data ingestion.
- Adversarial Rigor: Implementing "Red-Team" agents to stress-test system logic before deployment.
Kognia AI is a hierarchical MAS framework for autonomous research synthesis and strategic evaluation. It serves as the core proof-of-concept for my Nexus-Specialist architecture.
Below is a comprehensive log trace of the orchestration process. This "white-box" view provides key insights into the agentic lifecycle: agent handovers, tool invocations, and internal reasoning chains, facilitating deep debugging and system auditability.
I specialize in designing systems where specialized agents collaborate under a central orchestrator. Below is the blueprint I use for Autonomous Trade Reasoning:
I am currently engineering a Local Trade Reasoning Engine as a flagship demonstration of financial agentic frameworks.
- Orchestration: Engineering complex delegation loops with Google ADK.
- Memory & RAG: Leveraging Weaviate for hybrid search and long-term agentic memory.
- Local Inference: Optimizing Qwen3-30B-MoE models for high-performance local reasoning.
- Adversarial Logic: Developing a Risk Simulation Agent to veto setups based on market regime shifts.
- Observability: Implementing deep-trace logging to audit agentic "thought processes" in real-time.
- Agentic Frameworks: Google ADK, CrewAI, Hierarchical Orchestration, A2A Communication, MCP.
- Reasoning & RAG: Weaviate, Structured Output (Pydantic), Tool-Use (Function Calling), Langfuse, Long-Context Memory Management.
- Data & Backend: Python (
FastAPI,asyncio,pandas, numpy), Supabase (PostgreSQL), NoSQL (MongoDB), Render, Docker, REST/WebSocket API Integration. - Observability: Agent Evaluation, Technical Documentation, Systems Thinking.
I am open to collaborating on Applied AI and Agentic Orchestration projects or building bespoke autonomous workflows.
📫 Email: [email protected] 🔗 LinkedIn: linkedin.com/in/elikplim-kudowor