Home
Salome Monthe Chemiat - AI Researcher
I am an AI researcher at Strathmore University focused on reliable language model systems for high-stakes environments. My work investigates utility-aware retrieval, explicit safety constraints, and structured reasoning for trustworthy deployment. Research areas include neuro-symbolic AI, retrieval-augmented generation (RAG), and robust evaluation under real-world constraints.
Core research question: How can AI systems move beyond semantic similarity toward utility-aware and safety-constrained reasoning?
Short Bio
I hold an MSc in Information Technology (Distinction), with current research centered on LLM and RAG systems for maternal health decision support. My work analyzes failure modes including hallucination and reasoning inconsistency, and develops mitigation approaches based on symbolic constraints and retrieval redesign.
I also bring distributed systems experience to model evaluation and deployment, with a strong emphasis on reproducibility, reliability, and AI safety in constrained environments.
Research Snapshot
My research investigates why similarity-based retrieval often fails in clinical and high-risk tasks where the most semantically similar context is not always the safest or most useful. I design neuro-symbolic, utility-driven retrieval pipelines that enforce safety constraints and improve decision quality.
Current work focuses on maternal health applications and broader methods for reliable retrieval and generation in distributed AI systems. Read full research statement and projects.