AI Evaluation & Safety Attachee

About the role

Code Particles is looking for an intern to join its AI Evaluation & Safety team in Nairobi. The role covers designing evaluation suites for AI models and agents, red-teaming them against attacks such as prompt injection and jailbreaks, testing outputs that affect financial decisions for bias and fairness, and building automated evaluation pipelines integrated into CI/CD.

Key responsibilities

• Design and run evaluation suites for LLM features and agents before they ship
• Red-team models and agents by probing for prompt injection, jailbreaks and data leakage
• Test AI outputs affecting financial decisions for bias and fairness
• Build automated eval pipelines wired into CI/CD so model regressions block releases
• Curate golden datasets and behavioral regression suites for model changes
• Document model limitations and failure modes for compliance and audit trails
• Reproduce and dissect AI incidents reported from production
• Track quality metrics across model versions, prompts and agent configurations

Required profile

• Currently pursuing a degree in Computer Science, IT, Software Engineering, Statistics or a related field, with an industrial attachment required for academic credit
• Institution must provide an attachment letter and an evaluation framework (logbook, supervisor assessment or equivalent)
• An adversarial, detail-obsessed mindset and clear written communication, since findings become audit evidence
• Interest in AI safety, fairness and AI governance in finance
• Access to a reliable laptop and stable internet connection

Required skills

• Basic scripting in Python or JavaScript
• Understanding of the software development lifecycle and testing concepts
• Nice to have: testing frameworks (Playwright, Cypress, pytest), LLM eval tooling or benchmarks (promptfoo, HELM-style evals), basic statistics, CTF/bug bounty or security coursework, and knowledge of Kenya DPA, GDPR or financial regulation