Tokins Warmate

I've built and deployed AI systems from real client requirements, including RAG_OS, an agentic RAG system for Operating Systems concepts using LangChain, ChromaDB, HuggingFace embeddings, Groq-hosted Llama 3.1, and PostgreSQL.

While building RAG_OS, I identified a context-drift failure mode where retrieval returned source-consistent but semantically drifted answers. I designed a partitioned self-check and selective re-embedding mechanism that replaced full-index rebuilds, significantly reducing correction costs, while batch embedding and asynchronous handling kept the interface responsive.

As an Independent AI/ML Developer, I've delivered ensemble neonatal mortality prediction, genetic-algorithm image denoising, network intrusion monitoring, piracy-risk scoring, marine debris classification, and power-plant output optimization. My work has reached 87% accuracy for neonatal risk prediction, approximately 95% detection accuracy for intrusion monitoring, and a projected 10% reduction in power-plant operational costs.

Alongside my ongoing MSc in Computer Science and research work at Rivers State University, I presented applied AI to computing professionals as Chief Speaker at the Nigeria Computer Society's Rivers State Chapter seminar.