Sheriff Gatsi

I’m an AI Engineer building production-ready, AI-powered applications—especially LLM-driven agentic systems and practical customer service tools. I’m a recent Computer Engineering graduate, and I’ve combined research-minded work with hands-on engineering to ship real functionality.

In my AI Engineer internship (extended into an ongoing role) at CICOSY Zimbabwe, I designed and built customer service applications to automate responses and reduce manual query handling. I developed LLM-powered agentic systems using LangGraph to orchestrate multi-step reasoning and tool-calling workflows. I also built Python backend infrastructure to integrate AI systems with data sources and business logic.

My capstone work focused on privacy-preserving machine learning through federated learning and differential privacy. I designed a federated learning platform using layer-specific differential privacy with calibrated noise, and I integrated the Bulyan aggregation method to improve robustness against Byzantine and poisoning attacks.

I also build agentic LLM systems for data and support: an LLM-powered data analysis system that connects MySQL, PostgreSQL, APIs, and CSV files through LangGraph, and an e-commerce customer support agent using RAG to ground responses in product and policy data and reduce hallucinations.