Nosakhare Daniel Ahanor

Agentic AI and RAG workflow for automated cybersecurity threat and vulnerability assessment of SMEs' security posture. Includes an n8n data-ingestion pipeline that pulls cybersecurity news from Google Sheets, scrapes and validates content via an LLM agent, and indexes it into a RAG vector database.

Conversational AI security-awareness assistant for non-technical users. Responds instantly to security incidents with actionable advice via WhatsApp and Telegram.

Domain-adapted a foundation model (Llama 3.1 8B) via supervised fine-tuning and LoRA adapters so it can reason about phishing threats, not just classify them. Built an instruction-style data synthesis pipeline and a separate evaluation pipeline.

End-to-end MCP implementation with a FastMCP server exposing tools, prompts, and resources for arXiv paper research, a LangChain AI client, Docker containerization, CI/CD, and deployment to Render. Integrated with Claude Desktop, ChatGPT, and MCP Inspector.

SME-facing cybersecurity risk-assessment and monitoring app, published to Android and iOS as the Geiger App. Prototyped and researched at IIT (FHNW) as part of the EU Horizon 2020 GEIGER project, then owned and evolved into a production-ready application at cyberGEIGER GmbH. Includes a merge-triggered CI/CD orchestrator (GitHub Actions) that classifies conventional-commit history into release/patch/none and ships to environment-gated Play Store and Firebase targets with automatic semantic-version tagging.

Static web app for practicing Kubernetes and Cloud Native Associate (KCNA) exam questions. Randomized question selection with configurable domain distribution, a 60-second-per-question timer, instant score breakdown by domain, and a review of missed questions with explanations.