Emmanuel Watila

I’m an engineer and AI safety researcher. My work was usually set around product engineering, but has since shifted into empirical AI safety and alignment research: I build experiments that study how AI models behave when correctness, uncertainty, risks, and trust matter.

I’ve built Readlater, an AI-powered read-it-later platform for saving, organizing, searching, summarizing, and resurfacing important reading across a browser extension, web app, and MCP workflows.

Before formally completing a Technical AI Safety Cohort at BlueDot, I developed Krisis, a clinical LLM evaluation framework for measuring models behaviour (abstention, deferral, and uncertainty expression) in high-stakes medical reasoning tasks.

I was formally trained in Computer Engineering at the University of Maiduguri, where my thesis explored hybrid quantum-classical neural networks. Since then, I’ve worked across full-stack product engineering, open-source data tooling like breadroll, and technical writing.

Amongst my many interests are few I’m passionate about; neuroscience, physics (classical & quantum), and signal processing.

Doing

I’m currently researching AI Safety & Alignment (Supervised Fine-Tuning with Low-Rank Adaptation, Model Interpretability & Black Box Evaluation) and Quantum Machine Learning. Also learning Digital Signal Processing in Python.

Have a cool project?, feel free to hit me up.